AI Citation Patterns: How AI Engines Cite Sources (2026)
ChatGPT
Claude
Gemini
PerplexityAI engines cite sources in distinct, observable patterns: Perplexity uses numbered inline chips with a source list, ChatGPT shows inline numeric citations plus source bubbles, Google AI Overviews and AI Mode render inline links and a source carousel, Gemini lists sources beneath the answer, Microsoft Copilot uses footnote-style superscripts, and Claude.ai surfaces inline links when web search is enabled. Each pattern rewards a different content shape — extractable sentences, schema markup, and topical authority — so optimization is platform-specific.
TL;DR
ChatGPT cites a source in roughly 87% of web-grounded answers, Google AI Overviews in ~85%, Google AI Mode in ~76%, while Perplexity visits ~10 pages per query and surfaces 3-4 of them as numbered chips. To win citations across all of them, give every page one extractable answer, FAQPage / HowTo / Article schema, fresh dates, and a clear topical hub.
This page is part of the GEO reference hub and complements the AI Search Platform Comparison, What Is Source Selection?, and the deeper AI Citation Format Specification by Engine.
Definition
AI citation patterns are the structured, repeatable formats AI search engines use to attribute generated answers to underlying web sources. A citation pattern in this context is a combination of three things: the render style — how the citation appears in the user interface; the anchor mechanic — the specific claim, definition, list, or comparison that triggered the citation; and the selection unit — whether the engine cites a passage-level chunk, an entire URL, or a schema-extracted block.
Each major engine renders citations differently. Perplexity uses numbered inline chips paired with a sidebar source list. ChatGPT shows inline numeric markers plus a hover-revealed source bubble layer. Google AI Overviews and AI Mode embed inline hyperlinks alongside a horizontal source carousel. Gemini collapses sources beneath the answer behind a 'Sources' button. Microsoft Copilot relies on Bing-style footnote superscripts with a numbered list at the end. Claude.ai surfaces inline hyperlinks only when web search mode is active.
Citation patterns are distinct from training-data attribution. When ChatGPT or Claude answers without browsing, they draw on parameterized knowledge and do not cite. When they answer with web search enabled, they execute a retrieval pipeline (search → fetch → rank → extract → render) and produce the visible citation pattern documented on this page. Understanding the pattern per engine is the foundation of GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) — the optimization tactics that win a Perplexity numbered chip differ materially from those that win a Google AI Overviews source card or a ChatGPT source bubble.
Why Citation Patterns Matter
AI engines no longer return ten blue links — they synthesize an answer and then attach citations as evidence. The way each engine renders those citations changes three measurable outcomes:
- Whether users click through. Perplexity's numbered chips are denser and more click-intent than Gemini's collapsed source list. The same citation rate produces different referral traffic depending on render.
- What content gets selected. ChatGPT's source-bubble layer rewards short extractable claims; AIO's source carousel rewards Article + FAQPage schema; Copilot's footnotes reward H2/H3 heading hierarchies.
- How brand attribution works. Claude historically integrated sources implicitly; with web search on, Claude.ai now exposes inline links — so brand visibility on Claude is no longer purely about training-data presence.
Citation patterns also shape your measurement strategy. A 'citation share' metric that treats every citation equally will misweight Perplexity (high click-through, narrow source set) and Gemini (low click-through, broad source set). Sophisticated AI visibility programs weight citations by render type, query category, and click-through propensity — see AI Search KPIs for the full framework. Understanding the pattern per engine is also the first step in any AI visibility measurement program.
Citation Mechanics by Engine (2026)
| Platform | Render style | Citation rate | Click behavior |
|---|---|---|---|
| ChatGPT (web search on) | Inline numeric chips inside the answer + a source bubble layer (hover for title + URL) | ~87% of grounded answers | Click opens the source in a side panel or new tab |
| Perplexity | Numbered superscript chips inline, plus a numbered source list above and a horizontal source carousel | Cites ~3-4 sources from ~10 pages visited per query | Hover shows preview; click opens source |
| Google AI Overviews | Inline links within the answer + a horizontal source carousel of cards (favicon, title, domain) | ~85% of AIO answers cite at least one source | Click opens source in new tab; carousel shows alternates |
| Google AI Mode | Inline links + collapsible source list under the answer | ~76% of responses cite | Click opens source; long-tail heavy |
| Gemini | Source chips appear beneath the answer or behind a 'Sources' button | Varies by query type | Click opens source list |
| Microsoft Copilot (Bing-grounded) | Footnote-style superscripts inline + numbered list at the end | High for web-grounded answers | Click jumps to footnote then source |
| Claude.ai (web search on) | Inline hyperlinks within prose; sources panel on the right | Surfaces only when web search is invoked | Click opens source in new tab |
Citation rates are rolling averages reported across Averi (2026), Tinuiti Q1 2026, and Evertune (Mar 2026). They fluctuate by category and query type.
Each engine also differs in citation density — how many sources appear per answer. Perplexity averages 3-4 visible chips; ChatGPT averages 2-3 inline citations plus a deeper source bubble layer; AIO averages 1-3 source cards; Gemini and Claude average 1-2 inline links per answer. High-density engines reward exhaustive coverage pages; low-density engines reward a single dominant answer page.
Direct Citation vs Synthesized Mention vs Recommendation
Not every reference to your content in an AI answer is a citation. Three distinct attribution modes coexist, and conflating them is the most common analytics mistake teams make.
Direct citation. The engine renders an explicit, clickable reference — numbered chip, inline link, source card, or footnote — pointing to a specific URL. This is the only mode that drives measurable referral traffic and is what most AI visibility tools track.
Synthesized mention. The engine names a brand, product, or author in the prose without attaching a clickable source. This often happens when the engine pulls from training data, from low-confidence retrievals, or when multiple sources reinforce the same fact. Synthesized mentions still influence brand recall and can drive zero-click awareness, but they do not appear in referral analytics.
Recommendation. The engine actively suggests a brand, product, tool, or source as the answer to a recommendation-shaped query ('best X for Y', 'top 5 X'). Recommendations may or may not be citations — Perplexity tends to cite the recommended sources, while ChatGPT often recommends without citing.
| Mode | Render | Drives traffic? | Optimization lever |
|---|---|---|---|
| Direct citation | Numbered chip / inline link / source card | Yes | Extractable answers + schema + freshness |
| Synthesized mention | Plain-text brand mention | No (brand lift only) | Topical authority + entity associations + Wikipedia / Wikidata presence |
| Recommendation | 'We suggest [Brand]' or curated list | Sometimes | Comparative content + third-party validation + reviews on Reddit, G2, Trustpilot |
The same page can receive all three across different queries — for example, a SaaS comparison page might earn a direct citation on Perplexity, a synthesized mention on ChatGPT, and a recommendation on Gemini, all in the same week. Treat the three modes as a funnel: synthesized mentions and recommendations build the entity associations that make direct citations more likely later. A measurement framework that ignores synthesis and recommendations will systematically under-credit content that drives upstream brand awareness.
What Gets Cited: Chunks, Pages, and Snippet-Extracted Blocks
The selection unit varies by engine and dramatically affects content design. There are three observable selection units in 2026:
Chunks (passage-level). Most engines cite at the chunk level — a paragraph or 100-300-token window the engine extracted from your page. Perplexity, ChatGPT, and Google AI Overviews all operate this way. The implication: only the cited chunk matters for that citation; the rest of the page provides supporting topical context but rarely the visible quote.
Whole pages (URL-level). Some citations refer the user to a full page rather than a passage — typical when the answer summarizes structured content like a glossary entry, FAQ, or definitive guide. Google AI Mode and Gemini tend to cite this way more often, and the citation chip simply links to the canonical URL.
Snippet-extracted blocks (schema-driven). Engines that find FAQPage, HowTo, or Article JSON-LD often extract the structured block verbatim. This is the cleanest citation mechanic because the engine knows exactly what to lift, and the extracted block looks identical to what you marked up. FAQPage schema in particular has been linked to a ~2.7x citation lift in third-party studies, and HowTo schema to consistent extraction in AIO and Copilot.
For optimization, the practical takeaways are: write self-contained chunks (no 'as discussed above' cross-references), keep paragraphs tight (60-100 words), mark up Q&A and steps with schema, put the extractable answer near the top of each section rather than buried in the middle, and avoid splitting an answer across a heading boundary — the chunk extractor often stops at the next H2 or H3 and a half-answer below the boundary is invisible to the model.
Anchor Text Patterns
Anchor text — the visible text the engine uses for inline citations — follows predictable patterns across engines:
- Domain-only anchors. Perplexity numbered chips and ChatGPT source bubbles often surface only the domain on hover ('nytimes.com', 'stripe.com'). The implication: domain-level brand recognition matters more than headline cleverness.
- Title-based anchors. Google AI Overviews source cards display the page
and the domain favicon. Concise, keyword-led titles win; long branded titles get truncated.</li> <li>Numbered references. Perplexity, ChatGPT, and Microsoft Copilot all use numbered references ([1], [2], ¹, ²). The number itself is not addressable — engines map number → URL internally and the number is positional.</li> <li>Entity anchors. When the engine cites a brand or person rather than a URL, the anchor becomes the entity name (linked to Wikipedia, the brand site, or a Knowledge Graph card).</li> <li>Schema-derived anchors. When schema markup is used, the anchor sometimes mirrors a property in the markup (the name of a HowTo step, the headline of an Article).</li> </ul> <p>Optimize anchors by: keeping <title> under 60 characters and answer-led, ensuring the domain is short and brandable, registering the brand as an entity (Wikipedia, Wikidata, schema.org Organization with sameAs to authoritative profiles), and aligning JSON-LD headline / name properties with the extractable answer rather than with marketing copy.</p> <h2 id="practical-optimization-by-platform">Practical Optimization by Platform</h2> <h3 id="chatgpt">ChatGPT</h3> <ul> <li>Lead every page with a 1-2 sentence direct answer ChatGPT can quote.</li> <li>Maintain Article schema with datePublished and dateModified; ChatGPT favors freshness (Ahrefs reports AI assistants cite content that is on average ~25.7% newer than traditional search results).</li> <li>Use clean Markdown; serve .md mirrors when possible.</li> <li>Build topical clusters; ChatGPT's source bubble layer often surfaces sibling pages from the same domain.</li> </ul> <h3 id="perplexity">Perplexity</h3> <ul> <li>Optimize for inclusion in the ~10-page crawl set per query: structured data, clean canonicalization, fast TTFB.</li> <li>Pursue Reddit, YouTube, and LinkedIn presence — they account for ~25% of Perplexity's top citations.</li> <li>Provide a self-contained answer in the first paragraph; Perplexity's chips often pull a single sentence.</li> </ul> <h3 id="google-ai-overviews-and-ai-mode">Google AI Overviews and AI Mode</h3> <ul> <li>Capture featured-snippet-style answers (40-60-word definitions, list answers).</li> <li>Implement FAQPage and HowTo schema; AIO source cards heavily favor schema-rich pages.</li> <li>Maintain strong Core Web Vitals — AIO source selection is correlated with classic SEO authority.</li> </ul> <h3 id="gemini">Gemini</h3> <ul> <li>Gemini favors Reddit and authority publishers; create explicit comparison and pros/cons content that mirrors community discussion structure.</li> <li>Use Organization and Person schema with consistent sameAs to drive entity disambiguation.</li> </ul> <h3 id="microsoft-copilot">Microsoft Copilot</h3> <ul> <li>Copilot is Bing-grounded; ensure Bing Webmaster Tools indexing, sitemap submission, and IndexNow integration.</li> <li>Footnote-style citations reward lists and well-structured H2/H3 hierarchies.</li> </ul> <h3 id="claude">Claude</h3> <ul> <li>Claude.ai citations only appear when web search is enabled; for non-web sessions, brand presence in training data still matters.</li> <li>Optimize for clean prose with citable assertions — Claude tends to quote the cleanest paraphrasable sentence.</li> <li>See the dedicated Claude AI Citation Optimization guide for tactics.</li> </ul> <h2 id="examples-by-pattern">Examples by Pattern</h2> <p>The following examples illustrate each major citation pattern with hypothetical-but-typical query → answer → citation flows.</p> <p>Example 1 — Definitional query on Perplexity.</p> <p>Query: 'What is GEO?' → Perplexity returns a 2-paragraph answer with five numbered chips. Chip [1] points to a glossary entry; chips [2]-[5] point to longer guides. The cited glossary entry is a self-contained 80-word definition with FAQPage schema, and the chunk extracted into the answer is its first paragraph.</p> <p>Example 2 — How-to query on Google AI Overviews.</p> <p>Query: 'How to add llms.txt to a Next.js site?' → AIO renders a 5-step list with two source cards. Both sources have HowTo JSON-LD; the engine extracted the steps verbatim from the structured data, not from the prose. The source cards display each page's <title> and favicon.</p> <p>Example 3 — Comparison query on ChatGPT.</p> <p>Query: 'ChatGPT vs Perplexity for research' → ChatGPT returns a side-by-side table with inline numeric chips next to each row. The cited pages all use a comparison-table layout, so the engine could lift entire rows cleanly. The source bubbles, on hover, show only the domain.</p> <p>Example 4 — Recommendation query on Gemini.</p> <p>Query: 'Best SEO tools for AI search in 2026' → Gemini lists six tools without inline citations and shows a collapsed 'Sources' panel. The mentioned tools earn brand impressions but no referral traffic; tools with strong third-party reviews on Reddit and G2 dominated the list — a textbook synthesized-mention / recommendation pattern.</p> <p>Example 5 — Factual claim on Microsoft Copilot.</p> <p>Query: 'How many sources does ChatGPT cite per answer?' → Copilot returns the answer 'approximately 87% of web-grounded responses' with a footnote superscript ¹ leading to a Tinuiti research page. The cited page leads with the exact statistic in its first 150 words and uses Article schema.</p> <p>Example 6 — YMYL query on Claude.ai (web search on).</p> <p>Query: 'Latest CDC guidance on flu vaccine timing' → Claude returns a paragraph with two inline hyperlinks pointing to cdc.gov pages. Without web search enabled, Claude would have answered from training data with no clickable citations.</p> <p>Across all six examples, three patterns hold: extractable structure beats long prose, schema beats raw HTML, and freshness beats authority alone — the same principles that govern the wider AEO Content Checklist.</p> <h2 id="the-long-tail-reality">The Long-Tail Reality</h2> <p>Across every major engine, the majority of citations come from outside the top 20 most-cited sites. Evertune's Mar 2026 dataset shows:</p> <ul> <li>ChatGPT: top 3 sites combined = ~4.4% of citations; remaining sites = ~87.8%.</li> <li>Gemini: top 3 = ~3.2%; remaining = ~89.7%.</li> <li>Google AI Mode: top 3 = ~3.8%; remaining = ~87.4%.</li> <li>Google AI Overviews: top 3 = ~7.4%; remaining = ~83.4%.</li> <li>Perplexity: top 3 = ~24.9% (Reddit, YouTube, LinkedIn dominate); remaining = ~67.4%.</li> </ul> <p>The practical implication is that small and mid-size sites can win citations by owning specific, extractable answers — they are not locked out by Wikipedia or Forbes presence.</p> <h2 id="common-misconceptions">Common Misconceptions</h2> <ul> <li>'ChatGPT only uses footnotes.' Outdated. Since 2025, ChatGPT renders inline numeric chips and a source bubble overlay, not a footnote list.</li> <li>'Claude never cites sources.' Outdated. With web search on, Claude.ai exposes inline links; without web search, Claude still references training-data sources implicitly.</li> <li>'Citation rate is the same across categories.' False. Tinuiti Q1 2026 shows wide variation: AIO cites social media >4x more than Gemini, and YMYL categories (health, finance) have higher citation rates than retail.</li> <li>'You need top-3 ranking to be cited.' False. ~87% of ChatGPT citations come from sites outside the top 20.</li> <li>'Citations and rankings are the same thing.' False. Citation selection and traditional ranking share signals (authority, freshness, schema) but diverge on extractability and answer-shape — a page can rank #1 yet be uncited if no chunk is cleanly extractable.</li> </ul> <h2 id="tracking-your-citations">Tracking Your Citations</h2> <p>No single tool captures all engines, but a working stack in 2026 typically includes:</p> <ul> <li>Manual sampling — a weekly query bank of 20-50 questions across ChatGPT, Perplexity, AIO, AI Mode, Gemini, Copilot, and Claude.</li> <li>Citation tracking platforms — Profound, Otterly.AI, Peec AI, Siftly, Sight AI for automated coverage.</li> <li>GA4 referral channels — custom channel groupings using regex on chatgpt.com, chat.openai.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai referrers.</li> </ul> <p>See AI Search KPIs for the full measurement framework.</p> <h2 id="faq">FAQ</h2> <h3 id="q-which-platform-cites-sources-most-often">Q: Which platform cites sources most often?</h3> <p>ChatGPT (in web-grounded mode) and Google AI Overviews lead in 2026, citing roughly 85-87% of answers. Google AI Mode cites ~76% and Perplexity surfaces 3-4 sources per query from a ~10-page crawl. Rates fluctuate by category.</p> <h3 id="q-do-ai-citations-actually-drive-traffic">Q: Do AI citations actually drive traffic?</h3> <p>Yes, but unevenly. Perplexity and ChatGPT (with web search on) generate the highest click-through to cited sources; Gemini and Claude tend to convert citations into traffic at lower rates because their citation panels are collapsed by default.</p> <h3 id="q-how-is-citation-rate-measured">Q: How is citation rate measured?</h3> <p>The most common method is automated query sampling: send the same prompt set to each engine on a recurring schedule, parse the response for citation markup, and compute (queries with at least one citation) / (total queries). Tools like Profound and Siftly automate this; Tinuiti and Evertune publish category-level baselines.</p> <h3 id="q-can-you-force-an-ai-to-cite-your-page">Q: Can you force an AI to cite your page?</h3> <p>No. You can only raise the probability: serve clean Markdown, ship Article + FAQPage + HowTo schema, lead with extractable answers, keep dateModified fresh, build topical authority through internal linking, and earn third-party mentions in places AI engines crawl heavily (Reddit, Wikipedia, GitHub, mainstream press).</p> <h3 id="q-do-citation-patterns-differ-in-non-english-queries">Q: Do citation patterns differ in non-English queries?</h3> <p>Yes. Multilingual citation studies (see the related AI Search Multilingual Citation Patterns reference) show ChatGPT and Gemini default to English sources unless local-language signals are strong, while Perplexity tends to follow the user's query language more strictly.</p> <h3 id="q-are-citation-patterns-stable-or-do-they-change">Q: Are citation patterns stable, or do they change?</h3> <p>They change every release. Treat citation pattern reference pages as living documents reviewed at least quarterly; the version field on this page reflects when patterns were last verified end-to-end.</p> <h3 id="q-do-all-citations-carry-equal-weight">Q: Do all citations carry equal weight?</h3> <p>No. A Perplexity numbered chip in position [1] generates dramatically more click-through than a Gemini 'Sources' panel that requires expansion. Sophisticated AI visibility programs weight citations by render type, position, and query category — equal weighting flattens the signal.</p> <h3 id="q-should-i-optimize-for-citations-or-for-rankings">Q: Should I optimize for citations or for rankings?</h3> <p>Both. Citation selection and traditional ranking share upstream signals (authority, freshness, schema, page experience), but citation winners need an additional layer: extractable, self-contained chunks. A page that ranks #3 in Google can still be cited #1 in ChatGPT if its first paragraph is the cleanest available answer.</p> <h3 id="q-how-often-should-this-reference-be-reviewed">Q: How often should this reference be reviewed?</h3> <p>Every 90 days at minimum, and sooner whenever a major engine ships a new citation UI (Perplexity Pro, ChatGPT browse upgrades, Google Search Live, Gemini Deep Research). The render styles documented here are the most volatile section of the page; the underlying mechanics (chunks, pages, schema-extracted blocks) are stable.</p></div><div style="margin-top:64px"><h2 style="font-size:18px;font-weight:600;margin-bottom:16px;padding-bottom:8px;border-bottom:1px solid var(--border-default)">Related Articles</h2><div style="display:grid;grid-template-columns:repeat(auto-fill, minmax(280px, 1fr));gap:16px"><a style="text-decoration:none" data-related-article="ai-search-citation-types" data-related-position="1" href="/vi/geo/ai-search-citation-types"><div class="card"><span class="badge badge-green" style="margin-bottom:8px">reference</span><h3 style="font-size:15px;font-weight:600;color:var(--text-heading);margin-bottom:4px">AI Search Citation Types: How AI Attributes Sources</h3><p style="font-size:13px;color:var(--text-secondary);line-height:1.5">Reference for AI search citation types — inline, footnote, source card, attributed quote, implicit — with platform differences and how to optimize.</p></div></a><a style="text-decoration:none" data-related-article="what-is-source-selection" data-related-position="2" href="/vi/geo/what-is-source-selection"><div class="card"><span class="badge badge-green" style="margin-bottom:8px">definition</span><h3 style="font-size:15px;font-weight:600;color:var(--text-heading);margin-bottom:4px">What Is Source Selection in AI Search?</h3><p style="font-size:13px;color:var(--text-secondary);line-height:1.5">Source selection is how AI search engines evaluate, rank, and pick which web sources to cite when generating an answer. Learn what drives selection.</p></div></a><a style="text-decoration:none" data-related-article="ai-search-platform-comparison" data-related-position="3" href="/vi/reference/ai-search-platform-comparison"><div class="card"><span class="badge badge-green" style="margin-bottom:8px">comparison</span><h3 style="font-size:15px;font-weight:600;color:var(--text-heading);margin-bottom:4px">AI Search Platform Comparison</h3><p style="font-size:13px;color:var(--text-secondary);line-height:1.5">ChatGPT, Perplexity, AI Overviews, AI Mode, Claude, Copilot, and You.com compared: crawler UAs, citations, ranking signals, and per-platform GEO tactics.</p></div></a></div></div></article><div style="padding-left:40px" class="toc-sidebar"><nav style="position:sticky;top:88px;font-size:13px;line-height:1.6;max-height:calc(100vh - 100px);overflow-y:auto"><div style="font-family:var(--font-mono);font-size:10px;font-weight:600;text-transform:uppercase;letter-spacing:2px;color:var(--color-cool-gray);margin-bottom:12px">On this page</div><a href="#tl-dr" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">TL;DR</a><a href="#definition" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">Definition</a><a href="#why-citation-patterns-matter" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">Why Citation Patterns Matter</a><a href="#citation-mechanics-by-engine-2026" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">Citation Mechanics by Engine (2026)</a><a href="#direct-citation-vs-synthesized-mention-vs-recommendation" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">Direct Citation vs Synthesized Mention vs Recommendation</a><a href="#what-gets-cited-chunks-pages-and-snippet-extracted-blocks" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">What Gets Cited: Chunks, Pages, and Snippet-Extracted Blocks</a><a href="#anchor-text-patterns" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">Anchor Text Patterns</a><a href="#practical-optimization-by-platform" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">Practical Optimization by Platform</a><a href="#chatgpt" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">ChatGPT</a><a href="#perplexity" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Perplexity</a><a href="#google-ai-overviews-and-ai-mode" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Google AI Overviews and AI Mode</a><a href="#gemini" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Gemini</a><a href="#microsoft-copilot" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Microsoft Copilot</a><a href="#claude" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Claude</a><a href="#examples-by-pattern" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">Examples by Pattern</a><a href="#the-long-tail-reality" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">The Long-Tail Reality</a><a href="#common-misconceptions" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">Common Misconceptions</a><a href="#tracking-your-citations" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">Tracking Your Citations</a><a href="#faq" style="display:block;padding:4px 0;padding-left:0;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:none">FAQ</a><a href="#q-which-platform-cites-sources-most-often" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Q: Which platform cites sources most often?</a><a href="#q-do-ai-citations-actually-drive-traffic" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Q: Do AI citations actually drive traffic?</a><a href="#q-how-is-citation-rate-measured" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Q: How is citation rate measured?</a><a href="#q-can-you-force-an-ai-to-cite-your-page" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Q: Can you force an AI to cite your page?</a><a href="#q-do-citation-patterns-differ-in-non-english-queries" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Q: Do citation patterns differ in non-English queries?</a><a href="#q-are-citation-patterns-stable-or-do-they-change" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Q: Are citation patterns stable, or do they change?</a><a href="#q-do-all-citations-carry-equal-weight" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Q: Do all citations carry equal weight?</a><a href="#q-should-i-optimize-for-citations-or-for-rankings" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Q: Should I optimize for citations or for rankings?</a><a href="#q-how-often-should-this-reference-be-reviewed" style="display:block;padding:4px 0;padding-left:16px;color:var(--text-secondary);text-decoration:none;transition:color 0.15s ease;border-left:1px solid var(--border-default)">Q: How often should this reference be reviewed?</a></nav></div></div><style> @media (max-width: 1024px) { .toc-sidebar { display: none !important; } } @media (min-width: 1025px) { article { /* override grid for 3-col with TOC */ } } .article-tag-link:hover { border-color: var(--color-dark-green) !important; color: var(--text-primary) !important; background: rgba(0, 237, 100, 0.06) !important; } </style></main><section style="padding:64px 24px;background:var(--bg-surface);border-top:1px solid var(--border-subtle)"><div style="max-width:520px;margin:0 auto;text-align:center"><div style="font-family:var(--font-mono);font-size:10px;font-weight:600;text-transform:uppercase;letter-spacing:2px;color:var(--color-dark-green);margin-bottom:12px">Cập nhật tin tức</div><h2 style="font-family:var(--font-display);font-size:clamp(1.25rem, 3vw, 1.75rem);font-weight:400;color:var(--text-heading);margin-bottom:8px;line-height:1.3">Thông tin GEO & AI Search</h2><p style="font-size:14px;color:var(--text-secondary);line-height:1.7;margin-bottom:24px">Bài viết mới, cập nhật khung làm việc và phân tích ngành. Không spam, hủy đăng ký bất cứ lúc nào.</p><form style="display:flex;flex-wrap:wrap;gap:8px;max-width:420px;margin:0 auto"><input type="email" placeholder="email@cuaban.com" required="" style="flex:1;min-width:200px;padding:12px 16px;border-radius:var(--radius-pill);border:1px solid var(--border-default);background:var(--bg-page);color:var(--text-primary);font-family:var(--font-body);font-size:14px;outline:none;transition:border-color 0.2s ease" value=""/><button type="submit" class="btn btn-primary" style="padding:12px 24px;font-size:14px;opacity:1">Đăng ký</button></form></div></section><!--$--><!--/$--><footer style="background:var(--color-forest-black);border-top:1px solid var(--color-teal-gray);padding:64px 24px 32px"><div style="max-width:1200px;margin:0 auto"><div style="display:grid;grid-template-columns:repeat(auto-fit, minmax(180px, 1fr));gap:48px;margin-bottom:48px"><div><div style="margin-bottom:16px"><img alt="Geodocs.dev" loading="lazy" width="140" height="32" decoding="async" data-nimg="1" style="color:transparent;height:24px;width:auto" src="/geodocs-logo-dark.svg"/></div><p style="font-size:13px;color:var(--color-cool-gray);line-height:1.6;max-width:220px">Kiến thức có cấu trúc cho khả năng hiển thị tìm kiếm AI. Tài liệu tham khảo chuẩn cho GEO, AEO và tối ưu hóa tìm kiếm AI.</p></div><div><h4 style="font-family:var(--font-mono);font-size:11px;font-weight:600;text-transform:uppercase;letter-spacing:2px;color:var(--color-mongodb-green);margin-bottom:16px">Học</h4><ul style="list-style:none;display:flex;flex-direction:column;gap:10px"><li><a class="footer-link" href="/vi/geo/what-is-geo">What Is GEO?</a></li><li><a class="footer-link" href="/vi/aeo/what-is-aeo">What Is AEO?</a></li><li><a class="footer-link" href="/vi/geo/geo-vs-seo">GEO vs SEO</a></li><li><a class="footer-link" href="/vi/reference/geo-aeo-glossary">GEO Glossary</a></li></ul></div><div><h4 style="font-family:var(--font-mono);font-size:11px;font-weight:600;text-transform:uppercase;letter-spacing:2px;color:var(--color-mongodb-green);margin-bottom:16px">Xây dựng</h4><ul style="list-style:none;display:flex;flex-direction:column;gap:10px"><li><a class="footer-link" href="/vi/technical/llms-txt">llms.txt Reference</a></li><li><a class="footer-link" href="/vi/technical/how-to-create-llms-txt">Create llms.txt</a></li><li><a class="footer-link" href="/vi/technical/structured-data-for-ai-search">Structured Data</a></li><li><a class="footer-link" href="/vi/technical/ai-txt">ai.txt Reference</a></li></ul></div><div><h4 style="font-family:var(--font-mono);font-size:11px;font-weight:600;text-transform:uppercase;letter-spacing:2px;color:var(--color-mongodb-green);margin-bottom:16px">Chiến lược</h4><ul style="list-style:none;display:flex;flex-direction:column;gap:10px"><li><a class="footer-link" href="/vi/strategy/ai-visibility-measurement">AI Visibility</a></li><li><a class="footer-link" href="/vi/strategy/geo-content-strategy">Content Strategy</a></li><li><a class="footer-link" href="/vi/strategy/geo-roi-framework">GEO ROI</a></li><li><a class="footer-link" href="/vi/aeo/aeo-content-checklist">AEO Checklist</a></li></ul></div><div><h4 style="font-family:var(--font-mono);font-size:11px;font-weight:600;text-transform:uppercase;letter-spacing:2px;color:var(--color-mongodb-green);margin-bottom:16px">Tài nguyên</h4><ul style="list-style:none;display:flex;flex-direction:column;gap:10px"><li><a class="footer-link" href="https://github.com/Geodocs-dev">GitHub</a></li><li><a class="footer-link" href="/vi/contact">Contact</a></li><li><a class="footer-link" href="/vi/tags">Tags</a></li><li><a class="footer-link" href="/sitemap.xml">Sitemap</a></li><li><a class="footer-link" href="/llms.txt">llms.txt</a></li><li><a class="footer-link" href="/ai.txt">ai.txt</a></li></ul></div></div><div style="border-top:1px solid var(--color-teal-gray);padding-top:24px;display:flex;justify-content:space-between;align-items:center;flex-wrap:wrap;gap:16px"><p style="font-size:12px;color:var(--color-cool-gray)">© <!-- -->2026<!-- --> Geodocs.dev. <!-- -->Bảo lưu mọi quyền.</p><p style="font-size:12px;color:var(--color-cool-gray)"><a href="mailto:contact@geodocs.dev" class="footer-link">contact@geodocs.dev</a> · <!-- -->Xây dựng cho con người và AI agents.</p></div></div><style> .footer-link { font-size: 13px; color: var(--color-silver-teal); text-decoration: none; transition: color 0.2s ease; } .footer-link:hover { color: var(--color-white); } </style></footer><script> (function(){ var s=document.createElement('script'); s.src='https://cdn.jsdelivr.net/npm/mermaid@11/dist/mermaid.min.js'; s.defer=true; s.onload=function(){ var isDark=document.documentElement.getAttribute('data-theme')==='dark'; mermaid.initialize({startOnLoad:true,theme:isDark?'dark':'default',securityLevel:'loose'}); mermaid.run(); }; document.head.appendChild(s); })(); </script><script src="/_next/static/chunks/0_k5kz-r4593u.js" id="_R_" async=""></script><script>(self.__next_f=self.__next_f||[]).push([0])</script><script>self.__next_f.push([1,"1:\"$Sreact.fragment\"\n2:I[86402,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"GTMNoScript\"]\n3:I[59919,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"ThemeProvider\"]\n4:\"$Sreact.suspense\"\n5:I[86402,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"PostHogPageView\"]\n6:I[39756,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"default\"]\n7:I[37457,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"default\"]\n9:I[35264,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"AlgoliaSearchDialog\"]\na:I[56414,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"GeolifyAIDialog\"]\nb:I[86402,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"Analytics\"]\ne:I[97367,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"OutletBoundary\"]\n11:I[97367,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"ViewportBoundary\"]\n13:I[97367,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"MetadataBoundary\"]\n15:I[63491,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\",\"/_next/static/chunks/0z~59b-n8nq5n.js\"],\"default\"]\n:HL[\"/_next/static/chunks/0hbz95bqplbsv.css\",\"style\"]\n:HL[\"https://fonts.googleapis.com/css2?family=DM+Serif+Text\u0026family=Plus+Jakarta+Sans:wght@300;400;500;600;700\u0026family=Source+Code+Pro:wght@400;500;600\u0026display=swap\",\"style\"]\n"])</script><script>self.__next_f.push([1,"0:{\"P\":null,\"c\":[\"\",\"vi\",\"reference\",\"ai-citation-patterns\"],\"q\":\"\",\"i\":false,\"f\":[[[\"\",{\"children\":[[\"lang\",\"vi\",\"d\",null],{\"children\":[[\"section\",\"reference\",\"d\",null],{\"children\":[[\"slug\",\"ai-citation-patterns\",\"d\",null],{\"children\":[\"__PAGE__\",{}]}]}]}]},\"$undefined\",\"$undefined\",16],[[\"$\",\"$1\",\"c\",{\"children\":[[[\"$\",\"link\",\"0\",{\"rel\":\"stylesheet\",\"href\":\"/_next/static/chunks/0hbz95bqplbsv.css\",\"precedence\":\"next\",\"crossOrigin\":\"$undefined\",\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-0\",{\"src\":\"/_next/static/chunks/0ld4q8u-25eux.js\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-1\",{\"src\":\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-2\",{\"src\":\"/_next/static/chunks/0xf6s51ok42m_.js\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-3\",{\"src\":\"/_next/static/chunks/148t.fhegq9f1.js\",\"async\":true,\"nonce\":\"$undefined\"}],[\"$\",\"script\",\"script-4\",{\"src\":\"/_next/static/chunks/05cra..ka3fzk.js\",\"async\":true,\"nonce\":\"$undefined\"}]],[\"$\",\"html\",null,{\"lang\":\"en\",\"suppressHydrationWarning\":true,\"children\":[[\"$\",\"head\",null,{\"children\":[[\"$\",\"script\",null,{\"dangerouslySetInnerHTML\":{\"__html\":\"(function(){try{var t=localStorage.getItem('geodocs-theme');if(t==='dark'){document.documentElement.setAttribute('data-theme','dark')}else{document.documentElement.setAttribute('data-theme','light')}}catch(e){}})()\"}}],[\"$\",\"link\",null,{\"rel\":\"icon\",\"href\":\"/favicon.ico\",\"sizes\":\"any\"}],[\"$\",\"link\",null,{\"rel\":\"preconnect\",\"href\":\"https://fonts.googleapis.com\"}],[\"$\",\"link\",null,{\"rel\":\"preconnect\",\"href\":\"https://fonts.gstatic.com\",\"crossOrigin\":\"anonymous\"}],[\"$\",\"link\",null,{\"href\":\"https://fonts.googleapis.com/css2?family=DM+Serif+Text\u0026family=Plus+Jakarta+Sans:wght@300;400;500;600;700\u0026family=Source+Code+Pro:wght@400;500;600\u0026display=swap\",\"rel\":\"stylesheet\"}]]}],[\"$\",\"body\",null,{\"children\":[[\"$\",\"$L2\",null,{}],[\"$\",\"$L3\",null,{\"children\":[[\"$\",\"$4\",null,{\"fallback\":null,\"children\":[\"$\",\"$L5\",null,{}]}],[\"$\",\"$L6\",null,{\"parallelRouterKey\":\"children\",\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$L7\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":[\"$L8\",[]],\"forbidden\":\"$undefined\",\"unauthorized\":\"$undefined\"}],[\"$\",\"$L9\",null,{}],[\"$\",\"$La\",null,{}]]}],[\"$\",\"$Lb\",null,{}],[\"$\",\"script\",null,{\"dangerouslySetInnerHTML\":{\"__html\":\"\\n(function(){\\n var s=document.createElement('script');\\n s.src='https://cdn.jsdelivr.net/npm/mermaid@11/dist/mermaid.min.js';\\n s.defer=true;\\n s.onload=function(){\\n var isDark=document.documentElement.getAttribute('data-theme')==='dark';\\n mermaid.initialize({startOnLoad:true,theme:isDark?'dark':'default',securityLevel:'loose'});\\n mermaid.run();\\n };\\n document.head.appendChild(s);\\n})();\\n\"}}]]}]]}]]}],{\"children\":[[\"$\",\"$1\",\"c\",{\"children\":[[[\"$\",\"script\",\"script-0\",{\"src\":\"/_next/static/chunks/0-_gct68-k02d.js\",\"async\":true,\"nonce\":\"$undefined\"}]],\"$Lc\"]}],{\"children\":[[\"$\",\"$1\",\"c\",{\"children\":[null,[\"$\",\"$L6\",null,{\"parallelRouterKey\":\"children\",\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$L7\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":\"$undefined\",\"forbidden\":\"$undefined\",\"unauthorized\":\"$undefined\"}]]}],{\"children\":[[\"$\",\"$1\",\"c\",{\"children\":[null,[\"$\",\"$L6\",null,{\"parallelRouterKey\":\"children\",\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$L7\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":\"$undefined\",\"forbidden\":\"$undefined\",\"unauthorized\":\"$undefined\"}]]}],{\"children\":[[\"$\",\"$1\",\"c\",{\"children\":[\"$Ld\",[[\"$\",\"script\",\"script-0\",{\"src\":\"/_next/static/chunks/01avtxt88~n76.js\",\"async\":true,\"nonce\":\"$undefined\"}]],[\"$\",\"$Le\",null,{\"children\":[\"$\",\"$4\",null,{\"name\":\"Next.MetadataOutlet\",\"children\":\"$@f\"}]}]]}],{},null,false,null]},null,false,\"$@10\"]},null,false,\"$@10\"]},null,false,null]},null,false,null],[\"$\",\"$1\",\"h\",{\"children\":[null,[\"$\",\"$L11\",null,{\"children\":\"$L12\"}],[\"$\",\"div\",null,{\"hidden\":true,\"children\":[\"$\",\"$L13\",null,{\"children\":[\"$\",\"$4\",null,{\"name\":\"Next.Metadata\",\"children\":\"$L14\"}]}]}],null]}],false]],\"m\":\"$undefined\",\"G\":[\"$15\",[[\"$\",\"link\",\"0\",{\"rel\":\"stylesheet\",\"href\":\"/_next/static/chunks/0hbz95bqplbsv.css\",\"precedence\":\"next\",\"crossOrigin\":\"$undefined\",\"nonce\":\"$undefined\"}]]],\"S\":true,\"h\":null,\"s\":\"$undefined\",\"l\":\"$undefined\",\"p\":\"$undefined\",\"d\":\"$undefined\",\"b\":\"NIAK0ez3QwYT5WEUUjWPI\"}\n"])</script><script>self.__next_f.push([1,"16:[]\n10:\"$W16\"\n"])</script><script>self.__next_f.push([1,"17:I[2971,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\",\"/_next/static/chunks/0-_gct68-k02d.js\"],\"Header\"]\n"])</script><script>self.__next_f.push([1,"8:[[\"$\",\"$L17\",null,{\"lang\":\"en\",\"dict\":{\"nav\":{\"startHere\":\"Start Here\",\"geo\":\"GEO\",\"aeo\":\"AEO\",\"technical\":\"Technical\",\"strategy\":\"Strategy\",\"reference\":\"Reference\",\"tools\":\"Tools\",\"search\":\"Search docs...\",\"askGeolifyAI\":\"Ask GeolifyAI\",\"beta\":\"Beta\"},\"home\":{\"tagline\":\"The Canonical Knowledge System\",\"headline\":\"Structured knowledge for\",\"headlineHighlight\":\"AI search visibility\",\"subtitle\":\"GEO, AEO, and AI search optimization — defined, structured, and operationalized. Built for SEO professionals, developers, content teams, and AI agents.\",\"ctaPrimary\":\"Start with GEO\",\"ctaSecondary\":\"Read llms.txt spec\",\"trustSignals\":{\"answerFirst\":\"Answer-first content\",\"aiCitable\":\"AI-citable definitions\",\"machineReadable\":\"Machine-readable specs\",\"openFree\":\"Open \u0026 free\"},\"essentialReading\":\"Essential Reading\",\"geoPlaybooks\":\"GEO Playbooks\",\"playbooksSubtitle\":\"Canonical definitions, comparisons, and implementation guides. Start here to build your GEO foundation.\",\"readArticle\":\"Read article\",\"operatorsToolkit\":\"Operator's Toolkit\",\"technicalStandards\":\"Technical Standards for AI Search\",\"technicalSubtitle\":\"Implementation specs, file standards, and measurement frameworks. Everything you need to make your content AI-ready.\",\"knowledgeMap\":\"Knowledge Map\",\"browseBySection\":\"Browse by Section\",\"browseSubtitle\":\"Every concept has one canonical page. Explore the knowledge graph organized by domain.\",\"aiNativeDoc\":\"AI-Native Documentation\",\"builtForHumans\":\"Built for humans.\",\"readableByAI\":\"Readable by AI.\",\"aiNativeSubtitle\":\"Every page on geodocs.dev includes AI summary blocks, structured frontmatter, and machine-readable specs. Our content is designed to be cited by AI systems worldwide.\",\"aiContentSpec\":\"AI Content Spec\",\"viewLlmsTxt\":\"View llms.txt\"},\"sections\":{\"geo\":{\"title\":\"GEO - Generative Engine Optimization\",\"description\":\"Guides, definitions, and frameworks for optimizing content visibility in AI-generated answers.\"},\"aeo\":{\"title\":\"AEO - Answer Engine Optimization\",\"description\":\"How to structure content so AI systems can extract and cite direct answers.\"},\"technical\":{\"title\":\"Technical Implementation\",\"description\":\"llms.txt, ai.txt, structured data, and other technical specs for AI search readiness.\"},\"strategy\":{\"title\":\"Strategy \u0026 Frameworks\",\"description\":\"Business frameworks for AI search visibility — ROI, measurement, and content planning.\"},\"reference\":{\"title\":\"Reference\",\"description\":\"Glossary, cheatsheets, and canonical definitions for GEO/AEO terminology.\"},\"tools\":{\"title\":\"Tools \u0026 Platforms\",\"description\":\"Evaluations, comparisons, and stack recommendations for AI search optimization tools.\"},\"case-studies\":{\"title\":\"Case Studies\",\"description\":\"Evidence-based examples with real data — before/after results and industry applications.\"},\"ai-agents\":{\"title\":\"AI Agents\",\"description\":\"Machine-readable specs and documentation designed specifically for AI parsers and bots.\"}},\"article\":{\"minRead\":\"{min} min read\",\"words\":\"{count} words\",\"updated\":\"Updated {date}\",\"relatedArticles\":\"Related Articles\",\"onThisPage\":\"On this page\",\"topics\":\"Topics\",\"copy\":\"Copy\",\"copied\":\"Copied!\"},\"tags\":{\"browseByTopic\":\"Browse by Topic\",\"exploreTopics\":\"Explore {count} topics across all GEO, AEO, and AI search optimization articles.\",\"articlesTaggedWith\":\"{count} article tagged with \\\"{tag}\\\"|{count} articles tagged with \\\"{tag}\\\"\"},\"newsletter\":{\"stayUpdated\":\"Stay Updated\",\"title\":\"GEO \u0026 AI Search Insights\",\"description\":\"New articles, framework updates, and industry analysis. No spam, unsubscribe anytime.\",\"placeholder\":\"your@email.com\",\"subscribe\":\"Subscribe\",\"success\":\"You're subscribed! We'll keep you updated.\",\"error\":\"Something went wrong. Please try again.\"},\"footer\":{\"learn\":\"Learn\",\"build\":\"Build\",\"strategy\":\"Strategy\",\"resources\":\"Resources\",\"description\":\"Structured knowledge for AI search visibility. The canonical reference for GEO, AEO, and AI search optimization.\",\"allRightsReserved\":\"All rights reserved.\",\"builtForHumansAndAI\":\"Built for humans and AI agents.\"},\"common\":{\"articles\":\"articles\",\"article\":\"article\",\"page\":\"Page {current} of {total}\",\"noArticles\":\"No published articles in this section yet.\",\"checkBackSoon\":\"Content is being actively developed. Check back soon.\",\"loadingArticles\":\"Loading articles…\",\"prev\":\"← Prev\",\"next\":\"Next →\",\"goHome\":\"Go home\",\"startWithGEO\":\"Start with GEO\"},\"notFound\":{\"label\":\"404 - Page Not Found\",\"title\":\"This page doesn't exist yet.\",\"description\":\"The content you're looking for may be in development. Geodocs.dev is actively building the canonical knowledge base for GEO and AEO.\"},\"error\":{\"title\":\"Something went wrong\",\"description\":\"An unexpected error occurred. Our team has been notified.\",\"tryAgain\":\"Try Again\"},\"contact\":{\"getInTouch\":\"Get In Touch\",\"contactUs\":\"Contact Us\",\"heroDescription\":\"Have a question about GEO, AEO, or AI search optimization? Want to collaborate or contribute? We'd love to hear from you.\",\"name\":\"Name\",\"email\":\"Email\",\"subject\":\"Subject\",\"subjectPlaceholder\":\"What is this about?\",\"namePlaceholder\":\"Your name\",\"message\":\"Message\",\"messagePlaceholder\":\"Tell us more...\",\"sendMessage\":\"Send Message\",\"sending\":\"Sending...\",\"messageSent\":\"Message sent!\",\"thankYou\":\"Thank you for reaching out. We'll get back to you within 1-2 business days.\",\"sendAnother\":\"Send another message\",\"emailDirectly\":\"Or email us directly at\",\"validationName\":\"Please enter your name.\",\"validationEmail\":\"Please enter a valid email address.\",\"validationMessage\":\"Please enter a message (at least 10 characters).\",\"networkError\":\"Network error. Please try again.\",\"genericError\":\"Something went wrong. Please try again.\"},\"languagePicker\":{\"label\":\"Language\"}}}],\"$L18\",\"$L19\"]\n"])</script><script>self.__next_f.push([1,"1a:I[22016,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\",\"/_next/static/chunks/0-_gct68-k02d.js\",\"/_next/static/chunks/01avtxt88~n76.js\"],\"\"]\n1b:I[5500,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\",\"/_next/static/chunks/0-_gct68-k02d.js\"],\"Image\"]\n18:[\"$\",\"main\",null,{\"style\":{\"minHeight\":\"60vh\",\"display\":\"flex\",\"alignItems\":\"center\",\"justifyContent\":\"center\",\"textAlign\":\"center\",\"padding\":\"80px 24px\"},\"children\":[\"$\",\"div\",null,{\"children\":[[\"$\",\"div\",null,{\"style\":{\"fontFamily\":\"var(--font-mono)\",\"fontSize\":11,\"fontWeight\":600,\"textTransform\":\"uppercase\",\"letterSpacing\":3,\"color\":\"var(--color-cool-gray)\",\"marginBottom\":16},\"children\":\"404 - Page Not Found\"}],[\"$\",\"h1\",null,{\"style\":{\"fontFamily\":\"var(--font-display)\",\"fontSize\":\"clamp(2rem, 4vw, 3rem)\",\"fontWeight\":400,\"color\":\"var(--color-forest-black)\",\"marginBottom\":16},\"children\":\"This page doesn't exist yet.\"}],[\"$\",\"p\",null,{\"style\":{\"fontSize\":16,\"color\":\"var(--color-cool-gray)\",\"lineHeight\":1.7,\"maxWidth\":480,\"margin\":\"0 auto 32px\"},\"children\":\"The content you're looking for may be in development. Geodocs.dev is actively building the canonical knowledge base for GEO and AEO.\"}],[\"$\",\"div\",null,{\"style\":{\"display\":\"flex\",\"justifyContent\":\"center\",\"gap\":12},\"children\":[[\"$\",\"$L1a\",null,{\"href\":\"/\",\"className\":\"btn btn-primary\",\"children\":\"Go home\"}],[\"$\",\"$L1a\",null,{\"href\":\"/geo/what-is-geo\",\"className\":\"btn btn-outline\",\"children\":\"Start with GEO\"}]]}]]}]}]\n"])</script><script>self.__next_f.push([1,"19:[\"$\",\"footer\",null,{\"style\":{\"background\":\"var(--color-forest-black)\",\"borderTop\":\"1px solid var(--color-teal-gray)\",\"padding\":\"64px 24px 32px\"},\"children\":[[\"$\",\"div\",null,{\"style\":{\"maxWidth\":1200,\"margin\":\"0 auto\"},\"children\":[[\"$\",\"div\",null,{\"style\":{\"display\":\"grid\",\"gridTemplateColumns\":\"repeat(auto-fit, minmax(180px, 1fr))\",\"gap\":48,\"marginBottom\":48},\"children\":[[\"$\",\"div\",null,{\"children\":[[\"$\",\"div\",null,{\"style\":{\"marginBottom\":16},\"children\":[\"$\",\"$L1b\",null,{\"src\":\"/geodocs-logo-dark.svg\",\"alt\":\"Geodocs.dev\",\"width\":140,\"height\":32,\"style\":{\"height\":24,\"width\":\"auto\"}}]}],[\"$\",\"p\",null,{\"style\":{\"fontSize\":13,\"color\":\"var(--color-cool-gray)\",\"lineHeight\":1.6,\"maxWidth\":220},\"children\":\"Structured knowledge for AI search visibility. The canonical reference for GEO, AEO, and AI search optimization.\"}]]}],[[\"$\",\"div\",\"Learn\",{\"children\":[[\"$\",\"h4\",null,{\"style\":{\"fontFamily\":\"var(--font-mono)\",\"fontSize\":11,\"fontWeight\":600,\"textTransform\":\"uppercase\",\"letterSpacing\":2,\"color\":\"var(--color-mongodb-green)\",\"marginBottom\":16},\"children\":\"Learn\"}],[\"$\",\"ul\",null,{\"style\":{\"listStyle\":\"none\",\"display\":\"flex\",\"flexDirection\":\"column\",\"gap\":10},\"children\":[[\"$\",\"li\",\"/geo/what-is-geo\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/geo/what-is-geo\",\"className\":\"footer-link\",\"children\":\"What Is GEO?\"}]}],[\"$\",\"li\",\"/aeo/what-is-aeo\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/aeo/what-is-aeo\",\"className\":\"footer-link\",\"children\":\"What Is AEO?\"}]}],[\"$\",\"li\",\"/geo/geo-vs-seo\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/geo/geo-vs-seo\",\"className\":\"footer-link\",\"children\":\"GEO vs SEO\"}]}],[\"$\",\"li\",\"/reference/geo-aeo-glossary\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/reference/geo-aeo-glossary\",\"className\":\"footer-link\",\"children\":\"GEO Glossary\"}]}]]}]]}],[\"$\",\"div\",\"Build\",{\"children\":[[\"$\",\"h4\",null,{\"style\":{\"fontFamily\":\"var(--font-mono)\",\"fontSize\":11,\"fontWeight\":600,\"textTransform\":\"uppercase\",\"letterSpacing\":2,\"color\":\"var(--color-mongodb-green)\",\"marginBottom\":16},\"children\":\"Build\"}],[\"$\",\"ul\",null,{\"style\":{\"listStyle\":\"none\",\"display\":\"flex\",\"flexDirection\":\"column\",\"gap\":10},\"children\":[[\"$\",\"li\",\"/technical/llms-txt\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/technical/llms-txt\",\"className\":\"footer-link\",\"children\":\"llms.txt Reference\"}]}],[\"$\",\"li\",\"/technical/how-to-create-llms-txt\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/technical/how-to-create-llms-txt\",\"className\":\"footer-link\",\"children\":\"Create llms.txt\"}]}],[\"$\",\"li\",\"/technical/structured-data-for-ai-search\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/technical/structured-data-for-ai-search\",\"className\":\"footer-link\",\"children\":\"Structured Data\"}]}],[\"$\",\"li\",\"/technical/ai-txt\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/technical/ai-txt\",\"className\":\"footer-link\",\"children\":\"ai.txt Reference\"}]}]]}]]}],[\"$\",\"div\",\"Strategy\",{\"children\":[[\"$\",\"h4\",null,{\"style\":{\"fontFamily\":\"var(--font-mono)\",\"fontSize\":11,\"fontWeight\":600,\"textTransform\":\"uppercase\",\"letterSpacing\":2,\"color\":\"var(--color-mongodb-green)\",\"marginBottom\":16},\"children\":\"Strategy\"}],[\"$\",\"ul\",null,{\"style\":{\"listStyle\":\"none\",\"display\":\"flex\",\"flexDirection\":\"column\",\"gap\":10},\"children\":[[\"$\",\"li\",\"/strategy/ai-visibility-measurement\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/strategy/ai-visibility-measurement\",\"className\":\"footer-link\",\"children\":\"AI Visibility\"}]}],[\"$\",\"li\",\"/strategy/geo-content-strategy\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/strategy/geo-content-strategy\",\"className\":\"footer-link\",\"children\":\"Content Strategy\"}]}],[\"$\",\"li\",\"/strategy/geo-roi-framework\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/strategy/geo-roi-framework\",\"className\":\"footer-link\",\"children\":\"GEO ROI\"}]}],[\"$\",\"li\",\"/aeo/aeo-content-checklist\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/aeo/aeo-content-checklist\",\"className\":\"footer-link\",\"children\":\"AEO Checklist\"}]}]]}]]}],[\"$\",\"div\",\"Resources\",{\"children\":[[\"$\",\"h4\",null,{\"style\":{\"fontFamily\":\"var(--font-mono)\",\"fontSize\":11,\"fontWeight\":600,\"textTransform\":\"uppercase\",\"letterSpacing\":2,\"color\":\"var(--color-mongodb-green)\",\"marginBottom\":16},\"children\":\"Resources\"}],[\"$\",\"ul\",null,{\"style\":{\"listStyle\":\"none\",\"display\":\"flex\",\"flexDirection\":\"column\",\"gap\":10},\"children\":[[\"$\",\"li\",\"https://github.com/Geodocs-dev\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"https://github.com/Geodocs-dev\",\"className\":\"footer-link\",\"children\":\"GitHub\"}]}],[\"$\",\"li\",\"/contact\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/contact\",\"className\":\"footer-link\",\"children\":\"Contact\"}]}],[\"$\",\"li\",\"/tags\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/tags\",\"className\":\"footer-link\",\"children\":\"Tags\"}]}],[\"$\",\"li\",\"/sitemap.xml\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/sitemap.xml\",\"className\":\"footer-link\",\"children\":\"Sitemap\"}]}],\"$L1c\",\"$L1d\"]}]]}]]]}],\"$L1e\"]}],\"$L1f\"]}]\n"])</script><script>self.__next_f.push([1,"c:[[\"$\",\"script\",null,{\"dangerouslySetInnerHTML\":{\"__html\":\"document.documentElement.lang=\\\"vi\\\";\"}}],[\"$\",\"$L17\",null,{\"lang\":\"vi\",\"dict\":{\"nav\":{\"startHere\":\"Bắt đầu\",\"geo\":\"GEO\",\"aeo\":\"AEO\",\"technical\":\"Kỹ thuật\",\"strategy\":\"Chiến lược\",\"reference\":\"Tham khảo\",\"tools\":\"Công cụ\",\"search\":\"Tìm kiếm tài liệu...\",\"askGeolifyAI\":\"Hỏi GeolifyAI\",\"beta\":\"Beta\"},\"home\":{\"tagline\":\"Hệ thống kiến thức chuẩn\",\"headline\":\"Kiến thức có cấu trúc cho\",\"headlineHighlight\":\"khả năng hiển thị trên AI Search\",\"subtitle\":\"GEO, AEO và tối ưu hóa tìm kiếm AI — được định nghĩa, cấu trúc hóa và vận hành. Dành cho chuyên gia SEO, lập trình viên, đội ngũ nội dung và AI agents.\",\"ctaPrimary\":\"Bắt đầu với GEO\",\"ctaSecondary\":\"Đọc thông số llms.txt\",\"trustSignals\":{\"answerFirst\":\"Nội dung trả lời trực tiếp\",\"aiCitable\":\"Định nghĩa trích dẫn được bởi AI\",\"machineReadable\":\"Thông số đọc được bởi máy\",\"openFree\":\"Mở \u0026 miễn phí\"},\"essentialReading\":\"Đọc thiết yếu\",\"geoPlaybooks\":\"Cẩm nang GEO\",\"playbooksSubtitle\":\"Định nghĩa chuẩn, so sánh và hướng dẫn triển khai. Bắt đầu tại đây để xây dựng nền tảng GEO.\",\"readArticle\":\"Đọc bài viết\",\"operatorsToolkit\":\"Bộ công cụ vận hành\",\"technicalStandards\":\"Tiêu chuẩn kỹ thuật cho AI Search\",\"technicalSubtitle\":\"Thông số triển khai, chuẩn tệp và khung đo lường. Mọi thứ bạn cần để nội dung sẵn sàng cho AI.\",\"knowledgeMap\":\"Bản đồ kiến thức\",\"browseBySection\":\"Duyệt theo chủ đề\",\"browseSubtitle\":\"Mỗi khái niệm có một trang chuẩn. Khám phá đồ thị kiến thức được tổ chức theo lĩnh vực.\",\"aiNativeDoc\":\"Tài liệu AI-Native\",\"builtForHumans\":\"Xây dựng cho con người.\",\"readableByAI\":\"Đọc được bởi AI.\",\"aiNativeSubtitle\":\"Mỗi trang trên geodocs.dev bao gồm khối tóm tắt AI, frontmatter có cấu trúc và thông số đọc được bởi máy. Nội dung của chúng tôi được thiết kế để được trích dẫn bởi các hệ thống AI trên toàn thế giới.\",\"aiContentSpec\":\"Thông số nội dung AI\",\"viewLlmsTxt\":\"Xem llms.txt\"},\"sections\":{\"geo\":{\"title\":\"GEO - Tối ưu hóa công cụ tạo sinh\",\"description\":\"Hướng dẫn, định nghĩa và khung làm việc để tối ưu khả năng hiển thị nội dung trong câu trả lời do AI tạo ra.\"},\"aeo\":{\"title\":\"AEO - Tối ưu hóa công cụ trả lời\",\"description\":\"Cách cấu trúc nội dung để hệ thống AI có thể trích xuất và trích dẫn câu trả lời trực tiếp.\"},\"technical\":{\"title\":\"Triển khai kỹ thuật\",\"description\":\"llms.txt, ai.txt, dữ liệu có cấu trúc và các thông số kỹ thuật khác cho sự sẵn sàng tìm kiếm AI.\"},\"strategy\":{\"title\":\"Chiến lược \u0026 Khung làm việc\",\"description\":\"Khung kinh doanh cho khả năng hiển thị tìm kiếm AI — ROI, đo lường và lập kế hoạch nội dung.\"},\"reference\":{\"title\":\"Tham khảo\",\"description\":\"Bảng thuật ngữ, cheatsheet và định nghĩa chuẩn cho thuật ngữ GEO/AEO.\"},\"tools\":{\"title\":\"Công cụ \u0026 Nền tảng\",\"description\":\"Đánh giá, so sánh và khuyến nghị bộ công cụ tối ưu hóa tìm kiếm AI.\"},\"case-studies\":{\"title\":\"Nghiên cứu điển hình\",\"description\":\"Ví dụ dựa trên bằng chứng với dữ liệu thực — kết quả trước/sau và ứng dụng ngành.\"},\"ai-agents\":{\"title\":\"AI Agents\",\"description\":\"Thông số và tài liệu đọc được bởi máy, thiết kế riêng cho trình phân tích cú pháp và bot AI.\"}},\"article\":{\"minRead\":\"{min} phút đọc\",\"words\":\"{count} từ\",\"updated\":\"Cập nhật {date}\",\"relatedArticles\":\"Bài viết liên quan\",\"onThisPage\":\"Trên trang này\",\"topics\":\"Chủ đề\",\"copy\":\"Sao chép\",\"copied\":\"Đã sao chép!\"},\"tags\":{\"browseByTopic\":\"Duyệt theo chủ đề\",\"exploreTopics\":\"Khám phá {count} chủ đề trên tất cả bài viết về GEO, AEO và tối ưu hóa tìm kiếm AI.\",\"articlesTaggedWith\":\"{count} bài viết được gắn thẻ \\\"{tag}\\\"\"},\"newsletter\":{\"stayUpdated\":\"Cập nhật tin tức\",\"title\":\"Thông tin GEO \u0026 AI Search\",\"description\":\"Bài viết mới, cập nhật khung làm việc và phân tích ngành. Không spam, hủy đăng ký bất cứ lúc nào.\",\"placeholder\":\"email@cuaban.com\",\"subscribe\":\"Đăng ký\",\"success\":\"Bạn đã đăng ký! Chúng tôi sẽ cập nhật cho bạn.\",\"error\":\"Đã có lỗi xảy ra. Vui lòng thử lại.\"},\"footer\":{\"learn\":\"Học\",\"build\":\"Xây dựng\",\"strategy\":\"Chiến lược\",\"resources\":\"Tài nguyên\",\"description\":\"Kiến thức có cấu trúc cho khả năng hiển thị tìm kiếm AI. Tài liệu tham khảo chuẩn cho GEO, AEO và tối ưu hóa tìm kiếm AI.\",\"allRightsReserved\":\"Bảo lưu mọi quyền.\",\"builtForHumansAndAI\":\"Xây dựng cho con người và AI agents.\"},\"common\":{\"articles\":\"bài viết\",\"article\":\"bài viết\",\"page\":\"Trang {current}/{total}\",\"noArticles\":\"Chưa có bài viết nào được xuất bản trong mục này.\",\"checkBackSoon\":\"Nội dung đang được phát triển. Hãy quay lại sớm.\",\"loadingArticles\":\"Đang tải bài viết…\",\"prev\":\"← Trước\",\"next\":\"Tiếp →\",\"goHome\":\"Về trang chủ\",\"startWithGEO\":\"Bắt đầu với GEO\"},\"notFound\":{\"label\":\"404 - Không tìm thấy trang\",\"title\":\"Trang này chưa tồn tại.\",\"description\":\"Nội dung bạn đang tìm có thể đang được phát triển. Geodocs.dev đang tích cực xây dựng kho kiến thức chuẩn cho GEO và AEO.\"},\"error\":{\"title\":\"Đã xảy ra lỗi\",\"description\":\"Đã xảy ra lỗi không mong muốn. Đội ngũ của chúng tôi đã được thông báo.\",\"tryAgain\":\"Thử lại\"},\"contact\":{\"getInTouch\":\"Liên hệ\",\"contactUs\":\"Liên hệ với chúng tôi\",\"heroDescription\":\"Bạn có câu hỏi về GEO, AEO hoặc tối ưu hóa tìm kiếm AI? Muốn hợp tác hoặc đóng góp? Chúng tôi rất muốn nghe từ bạn.\",\"name\":\"Họ tên\",\"email\":\"Email\",\"subject\":\"Chủ đề\",\"subjectPlaceholder\":\"Nội dung liên hệ?\",\"namePlaceholder\":\"Tên của bạn\",\"message\":\"Tin nhắn\",\"messagePlaceholder\":\"Chia sẻ thêm...\",\"sendMessage\":\"Gửi tin nhắn\",\"sending\":\"Đang gửi...\",\"messageSent\":\"Đã gửi tin nhắn!\",\"thankYou\":\"Cảm ơn bạn đã liên hệ. Chúng tôi sẽ phản hồi trong 1-2 ngày làm việc.\",\"sendAnother\":\"Gửi tin nhắn khác\",\"emailDirectly\":\"Hoặc gửi email trực tiếp tới\",\"validationName\":\"Vui lòng nhập họ tên.\",\"validationEmail\":\"Vui lòng nhập địa chỉ email hợp lệ.\",\"validationMessage\":\"Vui lòng nhập tin nhắn (ít nhất 10 ký tự).\",\"networkError\":\"Lỗi mạng. Vui lòng thử lại.\",\"genericError\":\"Đã có lỗi xảy ra. Vui lòng thử lại.\"},\"languagePicker\":{\"label\":\"Ngôn ngữ\"}}}],\"$L20\",\"$L21\"]\n"])</script><script>self.__next_f.push([1,"28:I[43937,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\",\"/_next/static/chunks/0-_gct68-k02d.js\",\"/_next/static/chunks/01avtxt88~n76.js\"],\"ArticleTracker\"]\n1c:[\"$\",\"li\",\"/llms.txt\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/llms.txt\",\"className\":\"footer-link\",\"children\":\"llms.txt\"}]}]\n1d:[\"$\",\"li\",\"/ai.txt\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/ai.txt\",\"className\":\"footer-link\",\"children\":\"ai.txt\"}]}]\n1e:[\"$\",\"div\",null,{\"style\":{\"borderTop\":\"1px solid var(--color-teal-gray)\",\"paddingTop\":24,\"display\":\"flex\",\"justifyContent\":\"space-between\",\"alignItems\":\"center\",\"flexWrap\":\"wrap\",\"gap\":16},\"children\":[[\"$\",\"p\",null,{\"style\":{\"fontSize\":12,\"color\":\"var(--color-cool-gray)\"},\"children\":[\"© \",2026,\" Geodocs.dev. \",\"All rights reserved.\"]}],[\"$\",\"p\",null,{\"style\":{\"fontSize\":12,\"color\":\"var(--color-cool-gray)\"},\"children\":[[\"$\",\"a\",null,{\"href\":\"mailto:contact@geodocs.dev\",\"className\":\"footer-link\",\"children\":\"contact@geodocs.dev\"}],\" · \",\"Built for humans and AI agents.\"]}]]}]\n1f:[\"$\",\"style\",null,{\"children\":\"\\n .footer-link {\\n font-size: 13px;\\n color: var(--color-silver-teal);\\n text-decoration: none;\\n transition: color 0.2s ease;\\n }\\n .footer-link:hover {\\n color: var(--color-white);\\n }\\n \"}]\n20:[\"$\",\"$L6\",null,{\"parallelRouterKey\":\"children\",\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$L7\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":\"$undefined\",\"forbidden\":\"$undefined\",\"unauthorized\":\"$undefined\"}]\n"])</script><script>self.__next_f.push([1,"21:[\"$\",\"footer\",null,{\"style\":{\"background\":\"var(--color-forest-black)\",\"borderTop\":\"1px solid var(--color-teal-gray)\",\"padding\":\"64px 24px 32px\"},\"children\":[[\"$\",\"div\",null,{\"style\":{\"maxWidth\":1200,\"margin\":\"0 auto\"},\"children\":[[\"$\",\"div\",null,{\"style\":{\"display\":\"grid\",\"gridTemplateColumns\":\"repeat(auto-fit, minmax(180px, 1fr))\",\"gap\":48,\"marginBottom\":48},\"children\":[[\"$\",\"div\",null,{\"children\":[[\"$\",\"div\",null,{\"style\":{\"marginBottom\":16},\"children\":[\"$\",\"$L1b\",null,{\"src\":\"/geodocs-logo-dark.svg\",\"alt\":\"Geodocs.dev\",\"width\":140,\"height\":32,\"style\":{\"height\":24,\"width\":\"auto\"}}]}],[\"$\",\"p\",null,{\"style\":{\"fontSize\":13,\"color\":\"var(--color-cool-gray)\",\"lineHeight\":1.6,\"maxWidth\":220},\"children\":\"Kiến thức có cấu trúc cho khả năng hiển thị tìm kiếm AI. Tài liệu tham khảo chuẩn cho GEO, AEO và tối ưu hóa tìm kiếm AI.\"}]]}],[[\"$\",\"div\",\"Học\",{\"children\":[[\"$\",\"h4\",null,{\"style\":{\"fontFamily\":\"var(--font-mono)\",\"fontSize\":11,\"fontWeight\":600,\"textTransform\":\"uppercase\",\"letterSpacing\":2,\"color\":\"var(--color-mongodb-green)\",\"marginBottom\":16},\"children\":\"Học\"}],[\"$\",\"ul\",null,{\"style\":{\"listStyle\":\"none\",\"display\":\"flex\",\"flexDirection\":\"column\",\"gap\":10},\"children\":[[\"$\",\"li\",\"/vi/geo/what-is-geo\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/geo/what-is-geo\",\"className\":\"footer-link\",\"children\":\"What Is GEO?\"}]}],[\"$\",\"li\",\"/vi/aeo/what-is-aeo\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/aeo/what-is-aeo\",\"className\":\"footer-link\",\"children\":\"What Is AEO?\"}]}],[\"$\",\"li\",\"/vi/geo/geo-vs-seo\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/geo/geo-vs-seo\",\"className\":\"footer-link\",\"children\":\"GEO vs SEO\"}]}],[\"$\",\"li\",\"/vi/reference/geo-aeo-glossary\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/reference/geo-aeo-glossary\",\"className\":\"footer-link\",\"children\":\"GEO Glossary\"}]}]]}]]}],[\"$\",\"div\",\"Xây dựng\",{\"children\":[[\"$\",\"h4\",null,{\"style\":{\"fontFamily\":\"var(--font-mono)\",\"fontSize\":11,\"fontWeight\":600,\"textTransform\":\"uppercase\",\"letterSpacing\":2,\"color\":\"var(--color-mongodb-green)\",\"marginBottom\":16},\"children\":\"Xây dựng\"}],[\"$\",\"ul\",null,{\"style\":{\"listStyle\":\"none\",\"display\":\"flex\",\"flexDirection\":\"column\",\"gap\":10},\"children\":[[\"$\",\"li\",\"/vi/technical/llms-txt\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/technical/llms-txt\",\"className\":\"footer-link\",\"children\":\"llms.txt Reference\"}]}],[\"$\",\"li\",\"/vi/technical/how-to-create-llms-txt\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/technical/how-to-create-llms-txt\",\"className\":\"footer-link\",\"children\":\"Create llms.txt\"}]}],[\"$\",\"li\",\"/vi/technical/structured-data-for-ai-search\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/technical/structured-data-for-ai-search\",\"className\":\"footer-link\",\"children\":\"Structured Data\"}]}],[\"$\",\"li\",\"/vi/technical/ai-txt\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/technical/ai-txt\",\"className\":\"footer-link\",\"children\":\"ai.txt Reference\"}]}]]}]]}],[\"$\",\"div\",\"Chiến lược\",{\"children\":[[\"$\",\"h4\",null,{\"style\":{\"fontFamily\":\"var(--font-mono)\",\"fontSize\":11,\"fontWeight\":600,\"textTransform\":\"uppercase\",\"letterSpacing\":2,\"color\":\"var(--color-mongodb-green)\",\"marginBottom\":16},\"children\":\"Chiến lược\"}],[\"$\",\"ul\",null,{\"style\":{\"listStyle\":\"none\",\"display\":\"flex\",\"flexDirection\":\"column\",\"gap\":10},\"children\":[[\"$\",\"li\",\"/vi/strategy/ai-visibility-measurement\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/strategy/ai-visibility-measurement\",\"className\":\"footer-link\",\"children\":\"AI Visibility\"}]}],[\"$\",\"li\",\"/vi/strategy/geo-content-strategy\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/strategy/geo-content-strategy\",\"className\":\"footer-link\",\"children\":\"Content Strategy\"}]}],[\"$\",\"li\",\"/vi/strategy/geo-roi-framework\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/strategy/geo-roi-framework\",\"className\":\"footer-link\",\"children\":\"GEO ROI\"}]}],[\"$\",\"li\",\"/vi/aeo/aeo-content-checklist\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/aeo/aeo-content-checklist\",\"className\":\"footer-link\",\"children\":\"AEO Checklist\"}]}]]}]]}],[\"$\",\"div\",\"Tài nguyên\",{\"children\":[[\"$\",\"h4\",null,{\"style\":{\"fontFamily\":\"var(--font-mono)\",\"fontSize\":11,\"fontWeight\":600,\"textTransform\":\"uppercase\",\"letterSpacing\":2,\"color\":\"var(--color-mongodb-green)\",\"marginBottom\":16},\"children\":\"Tài nguyên\"}],[\"$\",\"ul\",null,{\"style\":{\"listStyle\":\"none\",\"display\":\"flex\",\"flexDirection\":\"column\",\"gap\":10},\"children\":[[\"$\",\"li\",\"https://github.com/Geodocs-dev\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"https://github.com/Geodocs-dev\",\"className\":\"footer-link\",\"children\":\"GitHub\"}]}],[\"$\",\"li\",\"/vi/contact\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/vi/contact\",\"className\":\"footer-link\",\"children\":\"Contact\"}]}],[\"$\",\"li\",\"/vi/tags\",{\"children\":\"$L22\"}],\"$L23\",\"$L24\",\"$L25\"]}]]}]]]}],\"$L26\"]}],\"$L27\"]}]\n"])</script><script>self.__next_f.push([1,"29:T52f,{\"@context\":\"https://schema.org\",\"@type\":\"TechArticle\",\"mainEntityOfPage\":{\"@type\":\"WebPage\",\"@id\":\"https://geodocs.dev/reference/ai-citation-patterns\"},\"headline\":\"AI Citation Patterns: How AI Engines Cite Sources (2026)\",\"description\":\"Reference of how ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Gemini, Microsoft Copilot, and Claude attribute sources in 2026 — with platform-specific optimization tactics.\",\"image\":\"https://geodocs.dev/geodocs-logo-dark.png\",\"url\":\"https://geodocs.dev/reference/ai-citation-patterns\",\"author\":{\"@type\":\"Person\",\"name\":\"Geodocs Research Team\"},\"publisher\":{\"@type\":\"Organization\",\"name\":\"Geodocs.dev\",\"logo\":{\"@type\":\"ImageObject\",\"url\":\"https://geodocs.dev/geodocs-logo-dark.png\"}},\"datePublished\":\"2025-04-25\",\"dateModified\":\"2026-05-01\",\"keywords\":\"AI citation patterns, AI source attribution, ChatGPT citation format, Perplexity citations, Google AI Overviews citations, how AI engines cite sources\",\"about\":[{\"@type\":\"Thing\",\"name\":\"AI Citation\"},{\"@type\":\"Thing\",\"name\":\"Source Attribution\"},{\"@type\":\"Thing\",\"name\":\"ChatGPT\"},{\"@type\":\"Thing\",\"name\":\"Perplexity\"},{\"@type\":\"Thing\",\"name\":\"Google AI Overviews\"},{\"@type\":\"Thing\",\"name\":\"Google AI Mode\"},{\"@type\":\"Thing\",\"name\":\"Gemini\"},{\"@type\":\"Thing\",\"name\":\"Microsoft Copilot\"},{\"@type\":\"Thing\",\"name\":\"Claude\"}]}2a:Te3d,"])</script><script>self.__next_f.push([1,"{\"@context\":\"https://schema.org\",\"@type\":\"FAQPage\",\"mainEntity\":[{\"@type\":\"Question\",\"name\":\"Which platform cites sources most often?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"ChatGPT (in web-grounded mode) and Google AI Overviews lead in 2026, citing roughly 85-87% of answers. Google AI Mode cites ~76% and Perplexity surfaces 3-4 sources per query from a ~10-page crawl. Rates fluctuate by category.\"}},{\"@type\":\"Question\",\"name\":\"Do AI citations actually drive traffic?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"Yes, but unevenly. Perplexity and ChatGPT (with web search on) generate the highest click-through to cited sources; Gemini and Claude tend to convert citations into traffic at lower rates because their citation panels are collapsed by default.\"}},{\"@type\":\"Question\",\"name\":\"How is citation rate measured?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"The most common method is automated query sampling: send the same prompt set to each engine on a recurring schedule, parse the response for citation markup, and compute (queries with at least one citation) / (total queries). Tools like Profound and Siftly automate this; Tinuiti and Evertune publish category-level baselines.\"}},{\"@type\":\"Question\",\"name\":\"Can you force an AI to cite your page?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"No. You can only raise the probability: serve clean Markdown, ship Article + FAQPage + HowTo schema, lead with extractable answers, keep dateModified fresh, build topical authority through internal linking, and earn third-party mentions in places AI engines crawl heavily (Reddit, Wikipedia, GitHub, mainstream press).\"}},{\"@type\":\"Question\",\"name\":\"Do citation patterns differ in non-English queries?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"Yes. Multilingual citation studies (see the related AI Search Multilingual Citation Patterns reference) show ChatGPT and Gemini default to English sources unless local-language signals are strong, while Perplexity tends to follow the user's query language more strictly.\"}},{\"@type\":\"Question\",\"name\":\"Are citation patterns stable, or do they change?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"They change every release. Treat citation pattern reference pages as living documents reviewed at least quarterly; the version field on this page reflects when patterns were last verified end-to-end.\"}},{\"@type\":\"Question\",\"name\":\"Do all citations carry equal weight?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"No. A Perplexity numbered chip in position [1] generates dramatically more click-through than a Gemini 'Sources' panel that requires expansion. Sophisticated AI visibility programs weight citations by render type, position, and query category — equal weighting flattens the signal.\"}},{\"@type\":\"Question\",\"name\":\"Should I optimize for citations or for rankings?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"Both. Citation selection and traditional ranking share upstream signals (authority, freshness, schema, page experience), but citation winners need an additional layer: extractable, self-contained chunks. A page that ranks #3 in Google can still be cited #1 in ChatGPT if its first paragraph is the cleanest available answer.\"}},{\"@type\":\"Question\",\"name\":\"How often should this reference be reviewed?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"Every 90 days at minimum, and sooner whenever a major engine ships a new citation UI (Perplexity Pro, ChatGPT browse upgrades, Google Search Live, Gemini Deep Research). The render styles documented here are the most volatile section of the page; the underlying mechanics (chunks, pages, schema-extracted blocks) are stable.\"}}]}"])</script><script>self.__next_f.push([1,"d:[[\"$\",\"main\",null,{\"style\":{\"minHeight\":\"70vh\"},\"children\":[[\"$\",\"$L28\",null,{\"meta\":{\"slug\":\"ai-citation-patterns\",\"section\":\"reference\",\"content_type\":\"reference\",\"difficulty\":\"intermediate\",\"primary_audience\":\"seo-specialist\",\"secondary_audiences\":[\"content-strategist\",\"marketer\",\"developer\"],\"word_count\":3235,\"reading_time_min\":17,\"has_code_snippet\":false,\"has_table\":false,\"citation_readiness\":\"reviewed\",\"series\":\"ai-citation-reference\",\"series_order\":1}}],[[\"$\",\"script\",null,{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"$29\"}}],[\"$\",\"script\",null,{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"{\\\"@context\\\":\\\"https://schema.org\\\",\\\"@type\\\":\\\"BreadcrumbList\\\",\\\"itemListElement\\\":[{\\\"@type\\\":\\\"ListItem\\\",\\\"position\\\":1,\\\"name\\\":\\\"Home\\\",\\\"item\\\":\\\"https://geodocs.dev\\\"},{\\\"@type\\\":\\\"ListItem\\\",\\\"position\\\":2,\\\"name\\\":\\\"REFERENCE\\\",\\\"item\\\":\\\"https://geodocs.dev/reference\\\"},{\\\"@type\\\":\\\"ListItem\\\",\\\"position\\\":3,\\\"name\\\":\\\"AI Citation Patterns: How AI Engines Cite Sources (2026)\\\",\\\"item\\\":\\\"https://geodocs.dev/reference/ai-citation-patterns\\\"}]}\"}}],[\"$\",\"script\",null,{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"$2a\"}}],\"$L2b\",\"$L2c\"]]}],\"$L2d\"]\n"])</script><script>self.__next_f.push([1,"2e:I[23150,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\",\"/_next/static/chunks/0-_gct68-k02d.js\",\"/_next/static/chunks/01avtxt88~n76.js\"],\"ShareButtons\"]\n2f:I[47654,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\",\"/_next/static/chunks/0-_gct68-k02d.js\",\"/_next/static/chunks/01avtxt88~n76.js\"],\"ExploreWithAI\"]\n33:I[11181,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\",\"/_next/static/chunks/0-_gct68-k02d.js\",\"/_next/static/chunks/01avtxt88~n76.js\"],\"Newsletter\"]\n22:[\"$\",\"$L1a\",null,{\"href\":\"/vi/tags\",\"className\":\"footer-link\",\"children\":\"Tags\"}]\n23:[\"$\",\"li\",\"/sitemap.xml\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/sitemap.xml\",\"className\":\"footer-link\",\"children\":\"Sitemap\"}]}]\n24:[\"$\",\"li\",\"/llms.txt\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/llms.txt\",\"className\":\"footer-link\",\"children\":\"llms.txt\"}]}]\n25:[\"$\",\"li\",\"/ai.txt\",{\"children\":[\"$\",\"$L1a\",null,{\"href\":\"/ai.txt\",\"className\":\"footer-link\",\"children\":\"ai.txt\"}]}]\n26:[\"$\",\"div\",null,{\"style\":{\"borderTop\":\"1px solid var(--color-teal-gray)\",\"paddingTop\":24,\"display\":\"flex\",\"justifyContent\":\"space-between\",\"alignItems\":\"center\",\"flexWrap\":\"wrap\",\"gap\":16},\"children\":[[\"$\",\"p\",null,{\"style\":{\"fontSize\":12,\"color\":\"var(--color-cool-gray)\"},\"children\":[\"© \",2026,\" Geodocs.dev. \",\"Bảo lưu mọi quyền.\"]}],[\"$\",\"p\",null,{\"style\":{\"fontSize\":12,\"color\":\"var(--color-cool-gray)\"},\"children\":[[\"$\",\"a\",null,{\"href\":\"mailto:contact@geodocs.dev\",\"className\":\"footer-link\",\"children\":\"contact@geodocs.dev\"}],\" · \",\"Xây dựng cho con người và AI agents.\"]}]]}]\n27:[\"$\",\"style\",null,{\"children\":\"\\n .footer-link {\\n font-size: 13px;\\n color: var(--color-silver-teal);\\n text-decoration: none;\\n transition: color 0.2s ease;\\n }\\n .footer-link:hover {\\n color: var(--color-white);\\n }\\n \"}]\n30:T5b39,"])</script><script>self.__next_f.push([1,"\u003cdiv class=\"ai-summary\"\u003e\u003cp\u003eAI engines cite sources in distinct, observable patterns: Perplexity uses numbered inline chips with a source list, ChatGPT shows inline numeric citations plus source bubbles, Google AI Overviews and AI Mode render inline links and a source carousel, Gemini lists sources beneath the answer, Microsoft Copilot uses footnote-style superscripts, and Claude.ai surfaces inline links when web search is enabled. Each pattern rewards a different content shape — extractable sentences, schema markup, and topical authority — so optimization is platform-specific.\u003c/p\u003e\u003c/div\u003e\n\u003ch3 id=\"tl-dr\"\u003eTL;DR\u003c/h3\u003e\n\u003cp\u003eChatGPT cites a source in roughly 87% of web-grounded answers, Google AI Overviews in ~85%, Google AI Mode in ~76%, while Perplexity visits ~10 pages per query and surfaces 3-4 of them as numbered chips. To win citations across all of them, give every page one extractable answer, FAQPage / HowTo / Article schema, fresh dates, and a clear topical hub.\u003c/p\u003e\n\u003cp\u003eThis page is part of the GEO reference hub and complements the AI Search Platform Comparison, What Is Source Selection?, and the deeper AI Citation Format Specification by Engine.\u003c/p\u003e\n\u003ch2 id=\"definition\"\u003eDefinition\u003c/h2\u003e\n\u003cp\u003eAI citation patterns are the structured, repeatable formats AI search engines use to attribute generated answers to underlying web sources. A citation pattern in this context is a combination of three things: the render style — how the citation appears in the user interface; the anchor mechanic — the specific claim, definition, list, or comparison that triggered the citation; and the selection unit — whether the engine cites a passage-level chunk, an entire URL, or a schema-extracted block.\u003c/p\u003e\n\u003cp\u003eEach major engine renders citations differently. Perplexity uses numbered inline chips paired with a sidebar source list. ChatGPT shows inline numeric markers plus a hover-revealed source bubble layer. Google AI Overviews and AI Mode embed inline hyperlinks alongside a horizontal source carousel. Gemini collapses sources beneath the answer behind a 'Sources' button. Microsoft Copilot relies on Bing-style footnote superscripts with a numbered list at the end. Claude.ai surfaces inline hyperlinks only when web search mode is active.\u003c/p\u003e\n\u003cp\u003eCitation patterns are distinct from training-data attribution. When ChatGPT or Claude answers without browsing, they draw on parameterized knowledge and do not cite. When they answer with web search enabled, they execute a retrieval pipeline (search → fetch → rank → extract → render) and produce the visible citation pattern documented on this page. Understanding the pattern per engine is the foundation of GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) — the optimization tactics that win a Perplexity numbered chip differ materially from those that win a Google AI Overviews source card or a ChatGPT source bubble.\u003c/p\u003e\n\u003ch2 id=\"why-citation-patterns-matter\"\u003eWhy Citation Patterns Matter\u003c/h2\u003e\n\u003cp\u003eAI engines no longer return ten blue links — they synthesize an answer and then attach citations as evidence. The way each engine renders those citations changes three measurable outcomes:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eWhether users click through. Perplexity's numbered chips are denser and more click-intent than Gemini's collapsed source list. The same citation rate produces different referral traffic depending on render.\u003c/li\u003e\n\u003cli\u003eWhat content gets selected. ChatGPT's source-bubble layer rewards short extractable claims; AIO's source carousel rewards Article + FAQPage schema; Copilot's footnotes reward H2/H3 heading hierarchies.\u003c/li\u003e\n\u003cli\u003eHow brand attribution works. Claude historically integrated sources implicitly; with web search on, Claude.ai now exposes inline links — so brand visibility on Claude is no longer purely about training-data presence.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eCitation patterns also shape your measurement strategy. A 'citation share' metric that treats every citation equally will misweight Perplexity (high click-through, narrow source set) and Gemini (low click-through, broad source set). Sophisticated AI visibility programs weight citations by render type, query category, and click-through propensity — see AI Search KPIs for the full framework. Understanding the pattern per engine is also the first step in any AI visibility measurement program.\u003c/p\u003e\n\u003ch2 id=\"citation-mechanics-by-engine-2026\"\u003eCitation Mechanics by Engine (2026)\u003c/h2\u003e\n\u003cdiv class=\"table-responsive\"\u003e\u003ctable\u003e\u003cthead\u003e\u003ctr\u003e\u003cth\u003ePlatform\u003c/th\u003e\u003cth\u003eRender style\u003c/th\u003e\u003cth\u003eCitation rate\u003c/th\u003e\u003cth\u003eClick behavior\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd\u003eChatGPT (web search on)\u003c/td\u003e\u003ctd\u003eInline numeric chips inside the answer + a source bubble layer (hover for title + URL)\u003c/td\u003e\u003ctd\u003e~87% of grounded answers\u003c/td\u003e\u003ctd\u003eClick opens the source in a side panel or new tab\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd\u003ePerplexity\u003c/td\u003e\u003ctd\u003eNumbered superscript chips inline, plus a numbered source list above and a horizontal source carousel\u003c/td\u003e\u003ctd\u003eCites ~3-4 sources from ~10 pages visited per query\u003c/td\u003e\u003ctd\u003eHover shows preview; click opens source\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd\u003eGoogle AI Overviews\u003c/td\u003e\u003ctd\u003eInline links within the answer + a horizontal source carousel of cards (favicon, title, domain)\u003c/td\u003e\u003ctd\u003e~85% of AIO answers cite at least one source\u003c/td\u003e\u003ctd\u003eClick opens source in new tab; carousel shows alternates\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd\u003eGoogle AI Mode\u003c/td\u003e\u003ctd\u003eInline links + collapsible source list under the answer\u003c/td\u003e\u003ctd\u003e~76% of responses cite\u003c/td\u003e\u003ctd\u003eClick opens source; long-tail heavy\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd\u003eGemini\u003c/td\u003e\u003ctd\u003eSource chips appear beneath the answer or behind a 'Sources' button\u003c/td\u003e\u003ctd\u003eVaries by query type\u003c/td\u003e\u003ctd\u003eClick opens source list\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd\u003eMicrosoft Copilot (Bing-grounded)\u003c/td\u003e\u003ctd\u003eFootnote-style superscripts inline + numbered list at the end\u003c/td\u003e\u003ctd\u003eHigh for web-grounded answers\u003c/td\u003e\u003ctd\u003eClick jumps to footnote then source\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd\u003eClaude.ai (web search on)\u003c/td\u003e\u003ctd\u003eInline hyperlinks within prose; sources panel on the right\u003c/td\u003e\u003ctd\u003eSurfaces only when web search is invoked\u003c/td\u003e\u003ctd\u003eClick opens source in new tab\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eCitation rates are rolling averages reported across Averi (2026), Tinuiti Q1 2026, and Evertune (Mar 2026). They fluctuate by category and query type.\u003c/p\u003e\n\u003cp\u003eEach engine also differs in citation density — how many sources appear per answer. Perplexity averages 3-4 visible chips; ChatGPT averages 2-3 inline citations plus a deeper source bubble layer; AIO averages 1-3 source cards; Gemini and Claude average 1-2 inline links per answer. High-density engines reward exhaustive coverage pages; low-density engines reward a single dominant answer page.\u003c/p\u003e\n\u003ch2 id=\"direct-citation-vs-synthesized-mention-vs-recommendation\"\u003eDirect Citation vs Synthesized Mention vs Recommendation\u003c/h2\u003e\n\u003cp\u003eNot every reference to your content in an AI answer is a citation. Three distinct attribution modes coexist, and conflating them is the most common analytics mistake teams make.\u003c/p\u003e\n\u003cp\u003eDirect citation. The engine renders an explicit, clickable reference — numbered chip, inline link, source card, or footnote — pointing to a specific URL. This is the only mode that drives measurable referral traffic and is what most AI visibility tools track.\u003c/p\u003e\n\u003cp\u003eSynthesized mention. The engine names a brand, product, or author in the prose without attaching a clickable source. This often happens when the engine pulls from training data, from low-confidence retrievals, or when multiple sources reinforce the same fact. Synthesized mentions still influence brand recall and can drive zero-click awareness, but they do not appear in referral analytics.\u003c/p\u003e\n\u003cp\u003eRecommendation. The engine actively suggests a brand, product, tool, or source as the answer to a recommendation-shaped query ('best X for Y', 'top 5 X'). Recommendations may or may not be citations — Perplexity tends to cite the recommended sources, while ChatGPT often recommends without citing.\u003c/p\u003e\n\u003cdiv class=\"table-responsive\"\u003e\u003ctable\u003e\u003cthead\u003e\u003ctr\u003e\u003cth\u003eMode\u003c/th\u003e\u003cth\u003eRender\u003c/th\u003e\u003cth\u003eDrives traffic?\u003c/th\u003e\u003cth\u003eOptimization lever\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd\u003eDirect citation\u003c/td\u003e\u003ctd\u003eNumbered chip / inline link / source card\u003c/td\u003e\u003ctd\u003eYes\u003c/td\u003e\u003ctd\u003eExtractable answers + schema + freshness\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd\u003eSynthesized mention\u003c/td\u003e\u003ctd\u003ePlain-text brand mention\u003c/td\u003e\u003ctd\u003eNo (brand lift only)\u003c/td\u003e\u003ctd\u003eTopical authority + entity associations + Wikipedia / Wikidata presence\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd\u003eRecommendation\u003c/td\u003e\u003ctd\u003e'We suggest [Brand]' or curated list\u003c/td\u003e\u003ctd\u003eSometimes\u003c/td\u003e\u003ctd\u003eComparative content + third-party validation + reviews on Reddit, G2, Trustpilot\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eThe same page can receive all three across different queries — for example, a SaaS comparison page might earn a direct citation on Perplexity, a synthesized mention on ChatGPT, and a recommendation on Gemini, all in the same week. Treat the three modes as a funnel: synthesized mentions and recommendations build the entity associations that make direct citations more likely later. A measurement framework that ignores synthesis and recommendations will systematically under-credit content that drives upstream brand awareness.\u003c/p\u003e\n\u003ch2 id=\"what-gets-cited-chunks-pages-and-snippet-extracted-blocks\"\u003eWhat Gets Cited: Chunks, Pages, and Snippet-Extracted Blocks\u003c/h2\u003e\n\u003cp\u003eThe selection unit varies by engine and dramatically affects content design. There are three observable selection units in 2026:\u003c/p\u003e\n\u003cp\u003eChunks (passage-level). Most engines cite at the chunk level — a paragraph or 100-300-token window the engine extracted from your page. Perplexity, ChatGPT, and Google AI Overviews all operate this way. The implication: only the cited chunk matters for that citation; the rest of the page provides supporting topical context but rarely the visible quote.\u003c/p\u003e\n\u003cp\u003eWhole pages (URL-level). Some citations refer the user to a full page rather than a passage — typical when the answer summarizes structured content like a glossary entry, FAQ, or definitive guide. Google AI Mode and Gemini tend to cite this way more often, and the citation chip simply links to the canonical URL.\u003c/p\u003e\n\u003cp\u003eSnippet-extracted blocks (schema-driven). Engines that find FAQPage, HowTo, or Article JSON-LD often extract the structured block verbatim. This is the cleanest citation mechanic because the engine knows exactly what to lift, and the extracted block looks identical to what you marked up. FAQPage schema in particular has been linked to a ~2.7x citation lift in third-party studies, and HowTo schema to consistent extraction in AIO and Copilot.\u003c/p\u003e\n\u003cp\u003eFor optimization, the practical takeaways are: write self-contained chunks (no 'as discussed above' cross-references), keep paragraphs tight (60-100 words), mark up Q\u0026A and steps with schema, put the extractable answer near the top of each section rather than buried in the middle, and avoid splitting an answer across a heading boundary — the chunk extractor often stops at the next H2 or H3 and a half-answer below the boundary is invisible to the model.\u003c/p\u003e\n\u003ch2 id=\"anchor-text-patterns\"\u003eAnchor Text Patterns\u003c/h2\u003e\n\u003cp\u003eAnchor text — the visible text the engine uses for inline citations — follows predictable patterns across engines:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eDomain-only anchors. Perplexity numbered chips and ChatGPT source bubbles often surface only the domain on hover ('nytimes.com', 'stripe.com'). The implication: domain-level brand recognition matters more than headline cleverness.\u003c/li\u003e\n\u003cli\u003eTitle-based anchors. Google AI Overviews source cards display the page \u003ctitle\u003e and the domain favicon. Concise, keyword-led titles win; long branded titles get truncated.\u003c/li\u003e\n\u003cli\u003eNumbered references. Perplexity, ChatGPT, and Microsoft Copilot all use numbered references ([1], [2], ¹, ²). The number itself is not addressable — engines map number → URL internally and the number is positional.\u003c/li\u003e\n\u003cli\u003eEntity anchors. When the engine cites a brand or person rather than a URL, the anchor becomes the entity name (linked to Wikipedia, the brand site, or a Knowledge Graph card).\u003c/li\u003e\n\u003cli\u003eSchema-derived anchors. When schema markup is used, the anchor sometimes mirrors a property in the markup (the name of a HowTo step, the headline of an Article).\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eOptimize anchors by: keeping \u003ctitle\u003e under 60 characters and answer-led, ensuring the domain is short and brandable, registering the brand as an entity (Wikipedia, Wikidata, schema.org Organization with sameAs to authoritative profiles), and aligning JSON-LD headline / name properties with the extractable answer rather than with marketing copy.\u003c/p\u003e\n\u003ch2 id=\"practical-optimization-by-platform\"\u003ePractical Optimization by Platform\u003c/h2\u003e\n\u003ch3 id=\"chatgpt\"\u003eChatGPT\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eLead every page with a 1-2 sentence direct answer ChatGPT can quote.\u003c/li\u003e\n\u003cli\u003eMaintain Article schema with datePublished and dateModified; ChatGPT favors freshness (Ahrefs reports AI assistants cite content that is on average ~25.7% newer than traditional search results).\u003c/li\u003e\n\u003cli\u003eUse clean Markdown; serve .md mirrors when possible.\u003c/li\u003e\n\u003cli\u003eBuild topical clusters; ChatGPT's source bubble layer often surfaces sibling pages from the same domain.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"perplexity\"\u003ePerplexity\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eOptimize for inclusion in the ~10-page crawl set per query: structured data, clean canonicalization, fast TTFB.\u003c/li\u003e\n\u003cli\u003ePursue Reddit, YouTube, and LinkedIn presence — they account for ~25% of Perplexity's top citations.\u003c/li\u003e\n\u003cli\u003eProvide a self-contained answer in the first paragraph; Perplexity's chips often pull a single sentence.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"google-ai-overviews-and-ai-mode\"\u003eGoogle AI Overviews and AI Mode\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eCapture featured-snippet-style answers (40-60-word definitions, list answers).\u003c/li\u003e\n\u003cli\u003eImplement FAQPage and HowTo schema; AIO source cards heavily favor schema-rich pages.\u003c/li\u003e\n\u003cli\u003eMaintain strong Core Web Vitals — AIO source selection is correlated with classic SEO authority.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"gemini\"\u003eGemini\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eGemini favors Reddit and authority publishers; create explicit comparison and pros/cons content that mirrors community discussion structure.\u003c/li\u003e\n\u003cli\u003eUse Organization and Person schema with consistent sameAs to drive entity disambiguation.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"microsoft-copilot\"\u003eMicrosoft Copilot\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eCopilot is Bing-grounded; ensure Bing Webmaster Tools indexing, sitemap submission, and IndexNow integration.\u003c/li\u003e\n\u003cli\u003eFootnote-style citations reward lists and well-structured H2/H3 hierarchies.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"claude\"\u003eClaude\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eClaude.ai citations only appear when web search is enabled; for non-web sessions, brand presence in training data still matters.\u003c/li\u003e\n\u003cli\u003eOptimize for clean prose with citable assertions — Claude tends to quote the cleanest paraphrasable sentence.\u003c/li\u003e\n\u003cli\u003eSee the dedicated Claude AI Citation Optimization guide for tactics.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch2 id=\"examples-by-pattern\"\u003eExamples by Pattern\u003c/h2\u003e\n\u003cp\u003eThe following examples illustrate each major citation pattern with hypothetical-but-typical query → answer → citation flows.\u003c/p\u003e\n\u003cp\u003eExample 1 — Definitional query on Perplexity.\u003c/p\u003e\n\u003cp\u003eQuery: 'What is GEO?' → Perplexity returns a 2-paragraph answer with five numbered chips. Chip [1] points to a glossary entry; chips [2]-[5] point to longer guides. The cited glossary entry is a self-contained 80-word definition with FAQPage schema, and the chunk extracted into the answer is its first paragraph.\u003c/p\u003e\n\u003cp\u003eExample 2 — How-to query on Google AI Overviews.\u003c/p\u003e\n\u003cp\u003eQuery: 'How to add llms.txt to a Next.js site?' → AIO renders a 5-step list with two source cards. Both sources have HowTo JSON-LD; the engine extracted the steps verbatim from the structured data, not from the prose. The source cards display each page's \u003ctitle\u003e and favicon.\u003c/p\u003e\n\u003cp\u003eExample 3 — Comparison query on ChatGPT.\u003c/p\u003e\n\u003cp\u003eQuery: 'ChatGPT vs Perplexity for research' → ChatGPT returns a side-by-side table with inline numeric chips next to each row. The cited pages all use a comparison-table layout, so the engine could lift entire rows cleanly. The source bubbles, on hover, show only the domain.\u003c/p\u003e\n\u003cp\u003eExample 4 — Recommendation query on Gemini.\u003c/p\u003e\n\u003cp\u003eQuery: 'Best SEO tools for AI search in 2026' → Gemini lists six tools without inline citations and shows a collapsed 'Sources' panel. The mentioned tools earn brand impressions but no referral traffic; tools with strong third-party reviews on Reddit and G2 dominated the list — a textbook synthesized-mention / recommendation pattern.\u003c/p\u003e\n\u003cp\u003eExample 5 — Factual claim on Microsoft Copilot.\u003c/p\u003e\n\u003cp\u003eQuery: 'How many sources does ChatGPT cite per answer?' → Copilot returns the answer 'approximately 87% of web-grounded responses' with a footnote superscript ¹ leading to a Tinuiti research page. The cited page leads with the exact statistic in its first 150 words and uses Article schema.\u003c/p\u003e\n\u003cp\u003eExample 6 — YMYL query on Claude.ai (web search on).\u003c/p\u003e\n\u003cp\u003eQuery: 'Latest CDC guidance on flu vaccine timing' → Claude returns a paragraph with two inline hyperlinks pointing to cdc.gov pages. Without web search enabled, Claude would have answered from training data with no clickable citations.\u003c/p\u003e\n\u003cp\u003eAcross all six examples, three patterns hold: extractable structure beats long prose, schema beats raw HTML, and freshness beats authority alone — the same principles that govern the wider AEO Content Checklist.\u003c/p\u003e\n\u003ch2 id=\"the-long-tail-reality\"\u003eThe Long-Tail Reality\u003c/h2\u003e\n\u003cp\u003eAcross every major engine, the majority of citations come from outside the top 20 most-cited sites. Evertune's Mar 2026 dataset shows:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eChatGPT: top 3 sites combined = ~4.4% of citations; remaining sites = ~87.8%.\u003c/li\u003e\n\u003cli\u003eGemini: top 3 = ~3.2%; remaining = ~89.7%.\u003c/li\u003e\n\u003cli\u003eGoogle AI Mode: top 3 = ~3.8%; remaining = ~87.4%.\u003c/li\u003e\n\u003cli\u003eGoogle AI Overviews: top 3 = ~7.4%; remaining = ~83.4%.\u003c/li\u003e\n\u003cli\u003ePerplexity: top 3 = ~24.9% (Reddit, YouTube, LinkedIn dominate); remaining = ~67.4%.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe practical implication is that small and mid-size sites can win citations by owning specific, extractable answers — they are not locked out by Wikipedia or Forbes presence.\u003c/p\u003e\n\u003ch2 id=\"common-misconceptions\"\u003eCommon Misconceptions\u003c/h2\u003e\n\u003cul\u003e\n\u003cli\u003e'ChatGPT only uses footnotes.' Outdated. Since 2025, ChatGPT renders inline numeric chips and a source bubble overlay, not a footnote list.\u003c/li\u003e\n\u003cli\u003e'Claude never cites sources.' Outdated. With web search on, Claude.ai exposes inline links; without web search, Claude still references training-data sources implicitly.\u003c/li\u003e\n\u003cli\u003e'Citation rate is the same across categories.' False. Tinuiti Q1 2026 shows wide variation: AIO cites social media \u003e4x more than Gemini, and YMYL categories (health, finance) have higher citation rates than retail.\u003c/li\u003e\n\u003cli\u003e'You need top-3 ranking to be cited.' False. ~87% of ChatGPT citations come from sites outside the top 20.\u003c/li\u003e\n\u003cli\u003e'Citations and rankings are the same thing.' False. Citation selection and traditional ranking share signals (authority, freshness, schema) but diverge on extractability and answer-shape — a page can rank #1 yet be uncited if no chunk is cleanly extractable.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch2 id=\"tracking-your-citations\"\u003eTracking Your Citations\u003c/h2\u003e\n\u003cp\u003eNo single tool captures all engines, but a working stack in 2026 typically includes:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eManual sampling — a weekly query bank of 20-50 questions across ChatGPT, Perplexity, AIO, AI Mode, Gemini, Copilot, and Claude.\u003c/li\u003e\n\u003cli\u003eCitation tracking platforms — Profound, Otterly.AI, Peec AI, Siftly, Sight AI for automated coverage.\u003c/li\u003e\n\u003cli\u003eGA4 referral channels — custom channel groupings using regex on chatgpt.com, chat.openai.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai referrers.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eSee AI Search KPIs for the full measurement framework.\u003c/p\u003e\n\u003ch2 id=\"faq\"\u003eFAQ\u003c/h2\u003e\n\u003ch3 id=\"q-which-platform-cites-sources-most-often\"\u003eQ: Which platform cites sources most often?\u003c/h3\u003e\n\u003cp\u003eChatGPT (in web-grounded mode) and Google AI Overviews lead in 2026, citing roughly 85-87% of answers. Google AI Mode cites ~76% and Perplexity surfaces 3-4 sources per query from a ~10-page crawl. Rates fluctuate by category.\u003c/p\u003e\n\u003ch3 id=\"q-do-ai-citations-actually-drive-traffic\"\u003eQ: Do AI citations actually drive traffic?\u003c/h3\u003e\n\u003cp\u003eYes, but unevenly. Perplexity and ChatGPT (with web search on) generate the highest click-through to cited sources; Gemini and Claude tend to convert citations into traffic at lower rates because their citation panels are collapsed by default.\u003c/p\u003e\n\u003ch3 id=\"q-how-is-citation-rate-measured\"\u003eQ: How is citation rate measured?\u003c/h3\u003e\n\u003cp\u003eThe most common method is automated query sampling: send the same prompt set to each engine on a recurring schedule, parse the response for citation markup, and compute (queries with at least one citation) / (total queries). Tools like Profound and Siftly automate this; Tinuiti and Evertune publish category-level baselines.\u003c/p\u003e\n\u003ch3 id=\"q-can-you-force-an-ai-to-cite-your-page\"\u003eQ: Can you force an AI to cite your page?\u003c/h3\u003e\n\u003cp\u003eNo. You can only raise the probability: serve clean Markdown, ship Article + FAQPage + HowTo schema, lead with extractable answers, keep dateModified fresh, build topical authority through internal linking, and earn third-party mentions in places AI engines crawl heavily (Reddit, Wikipedia, GitHub, mainstream press).\u003c/p\u003e\n\u003ch3 id=\"q-do-citation-patterns-differ-in-non-english-queries\"\u003eQ: Do citation patterns differ in non-English queries?\u003c/h3\u003e\n\u003cp\u003eYes. Multilingual citation studies (see the related AI Search Multilingual Citation Patterns reference) show ChatGPT and Gemini default to English sources unless local-language signals are strong, while Perplexity tends to follow the user's query language more strictly.\u003c/p\u003e\n\u003ch3 id=\"q-are-citation-patterns-stable-or-do-they-change\"\u003eQ: Are citation patterns stable, or do they change?\u003c/h3\u003e\n\u003cp\u003eThey change every release. Treat citation pattern reference pages as living documents reviewed at least quarterly; the version field on this page reflects when patterns were last verified end-to-end.\u003c/p\u003e\n\u003ch3 id=\"q-do-all-citations-carry-equal-weight\"\u003eQ: Do all citations carry equal weight?\u003c/h3\u003e\n\u003cp\u003eNo. A Perplexity numbered chip in position [1] generates dramatically more click-through than a Gemini 'Sources' panel that requires expansion. Sophisticated AI visibility programs weight citations by render type, position, and query category — equal weighting flattens the signal.\u003c/p\u003e\n\u003ch3 id=\"q-should-i-optimize-for-citations-or-for-rankings\"\u003eQ: Should I optimize for citations or for rankings?\u003c/h3\u003e\n\u003cp\u003eBoth. Citation selection and traditional ranking share upstream signals (authority, freshness, schema, page experience), but citation winners need an additional layer: extractable, self-contained chunks. A page that ranks #3 in Google can still be cited #1 in ChatGPT if its first paragraph is the cleanest available answer.\u003c/p\u003e\n\u003ch3 id=\"q-how-often-should-this-reference-be-reviewed\"\u003eQ: How often should this reference be reviewed?\u003c/h3\u003e\n\u003cp\u003eEvery 90 days at minimum, and sooner whenever a major engine ships a new citation UI (Perplexity Pro, ChatGPT browse upgrades, Google Search Live, Gemini Deep Research). The render styles documented here are the most volatile section of the page; the underlying mechanics (chunks, pages, schema-extracted blocks) are stable.\u003c/p\u003e"])</script><script>self.__next_f.push([1,"2b:[\"$\",\"div\",null,{\"style\":{\"display\":\"grid\",\"gridTemplateColumns\":\"1fr min(var(--content-max-width), 100%) 1fr\",\"gap\":0,\"maxWidth\":1200,\"margin\":\"0 auto\",\"padding\":\"48px 24px 80px\"},\"children\":[[\"$\",\"div\",null,{}],[\"$\",\"article\",null,{\"children\":[[\"$\",\"h1\",null,{\"style\":{\"fontFamily\":\"var(--font-display)\",\"fontSize\":\"clamp(1.75rem, 4vw, 2.5rem)\",\"fontWeight\":400,\"lineHeight\":1.2,\"color\":\"var(--text-heading)\",\"marginBottom\":16},\"children\":\"AI Citation Patterns: How AI Engines Cite Sources (2026)\"}],[\"$\",\"div\",null,{\"className\":\"metadata-bar\",\"style\":{\"marginBottom\":24},\"children\":[[\"$\",\"span\",null,{\"className\":\"metadata-bar__tag metadata-bar__tag--section\",\"children\":\"reference\"}],[\"$\",\"span\",null,{\"className\":\"metadata-bar__tag\",\"children\":\"reference\"}],[\"$\",\"span\",null,{\"className\":\"metadata-bar__tag\",\"children\":\"intermediate\"}],[\"$\",\"span\",null,{\"className\":\"metadata-bar__dot\"}],[\"$\",\"span\",null,{\"children\":[17,\" min read\"]}],[\"$\",\"span\",null,{\"className\":\"metadata-bar__dot\"}],[\"$\",\"span\",null,{\"children\":[\"3,235\",\" words\"]}],[[\"$\",\"span\",null,{\"className\":\"metadata-bar__dot\"}],[\"$\",\"span\",null,{\"children\":[\"Updated\",\" \",\"May 2026\"]}]]]}],[\"$\",\"$L2e\",null,{\"title\":\"AI Citation Patterns: How AI Engines Cite Sources (2026)\",\"url\":\"https://geodocs.dev/reference/ai-citation-patterns\"}],[\"$\",\"$L2f\",null,{\"title\":\"AI Citation Patterns: How AI Engines Cite Sources (2026)\",\"url\":\"https://geodocs.dev/reference/ai-citation-patterns\",\"section\":\"reference\",\"slug\":\"ai-citation-patterns\"}],[\"$\",\"div\",null,{\"className\":\"prose\",\"dangerouslySetInnerHTML\":{\"__html\":\"$30\"}}],\"$L31\",false]}],\"$L32\"]}]\n"])</script><script>self.__next_f.push([1,"2c:[\"$\",\"style\",null,{\"children\":\"\\n @media (max-width: 1024px) {\\n .toc-sidebar { display: none !important; }\\n }\\n @media (min-width: 1025px) {\\n article {\\n /* override grid for 3-col with TOC */\\n }\\n }\\n .article-tag-link:hover {\\n border-color: var(--color-dark-green) !important;\\n color: var(--text-primary) !important;\\n background: rgba(0, 237, 100, 0.06) !important;\\n }\\n \"}]\n2d:[\"$\",\"$L33\",null,{\"dict\":\"$c:1:props:dict:newsletter\"}]\n"])</script><script>self.__next_f.push([1,"31:[\"$\",\"div\",null,{\"style\":{\"marginTop\":64},\"children\":[[\"$\",\"h2\",null,{\"style\":{\"fontSize\":18,\"fontWeight\":600,\"marginBottom\":16,\"paddingBottom\":8,\"borderBottom\":\"1px solid var(--border-default)\"},\"children\":\"Related Articles\"}],[\"$\",\"div\",null,{\"style\":{\"display\":\"grid\",\"gridTemplateColumns\":\"repeat(auto-fill, minmax(280px, 1fr))\",\"gap\":16},\"children\":[[\"$\",\"$L1a\",\"ai-search-citation-types\",{\"href\":\"/vi/geo/ai-search-citation-types\",\"style\":{\"textDecoration\":\"none\"},\"data-related-article\":\"ai-search-citation-types\",\"data-related-position\":1,\"children\":[\"$\",\"div\",null,{\"className\":\"card\",\"children\":[[\"$\",\"span\",null,{\"className\":\"badge badge-green\",\"style\":{\"marginBottom\":8},\"children\":\"reference\"}],[\"$\",\"h3\",null,{\"style\":{\"fontSize\":15,\"fontWeight\":600,\"color\":\"var(--text-heading)\",\"marginBottom\":4},\"children\":\"AI Search Citation Types: How AI Attributes Sources\"}],[\"$\",\"p\",null,{\"style\":{\"fontSize\":13,\"color\":\"var(--text-secondary)\",\"lineHeight\":1.5},\"children\":\"Reference for AI search citation types — inline, footnote, source card, attributed quote, implicit — with platform differences and how to optimize.\"}]]}]}],[\"$\",\"$L1a\",\"what-is-source-selection\",{\"href\":\"/vi/geo/what-is-source-selection\",\"style\":{\"textDecoration\":\"none\"},\"data-related-article\":\"what-is-source-selection\",\"data-related-position\":2,\"children\":[\"$\",\"div\",null,{\"className\":\"card\",\"children\":[[\"$\",\"span\",null,{\"className\":\"badge badge-green\",\"style\":{\"marginBottom\":8},\"children\":\"definition\"}],[\"$\",\"h3\",null,{\"style\":{\"fontSize\":15,\"fontWeight\":600,\"color\":\"var(--text-heading)\",\"marginBottom\":4},\"children\":\"What Is Source Selection in AI Search?\"}],[\"$\",\"p\",null,{\"style\":{\"fontSize\":13,\"color\":\"var(--text-secondary)\",\"lineHeight\":1.5},\"children\":\"Source selection is how AI search engines evaluate, rank, and pick which web sources to cite when generating an answer. Learn what drives selection.\"}]]}]}],[\"$\",\"$L1a\",\"ai-search-platform-comparison\",{\"href\":\"/vi/reference/ai-search-platform-comparison\",\"style\":{\"textDecoration\":\"none\"},\"data-related-article\":\"ai-search-platform-comparison\",\"data-related-position\":3,\"children\":[\"$\",\"div\",null,{\"className\":\"card\",\"children\":[[\"$\",\"span\",null,{\"className\":\"badge badge-green\",\"style\":{\"marginBottom\":8},\"children\":\"comparison\"}],[\"$\",\"h3\",null,{\"style\":{\"fontSize\":15,\"fontWeight\":600,\"color\":\"var(--text-heading)\",\"marginBottom\":4},\"children\":\"AI Search Platform Comparison\"}],[\"$\",\"p\",null,{\"style\":{\"fontSize\":13,\"color\":\"var(--text-secondary)\",\"lineHeight\":1.5},\"children\":\"ChatGPT, Perplexity, AI Overviews, AI Mode, Claude, Copilot, and You.com compared: crawler UAs, citations, ranking signals, and per-platform GEO tactics.\"}]]}]}]]}]]}]\n"])</script><script>self.__next_f.push([1,"32:[\"$\",\"div\",null,{\"style\":{\"paddingLeft\":40},\"className\":\"toc-sidebar\",\"children\":[\"$\",\"nav\",null,{\"style\":{\"position\":\"sticky\",\"top\":88,\"fontSize\":13,\"lineHeight\":1.6,\"maxHeight\":\"calc(100vh - 100px)\",\"overflowY\":\"auto\"},\"children\":[[\"$\",\"div\",null,{\"style\":{\"fontFamily\":\"var(--font-mono)\",\"fontSize\":10,\"fontWeight\":600,\"textTransform\":\"uppercase\",\"letterSpacing\":2,\"color\":\"var(--color-cool-gray)\",\"marginBottom\":12},\"children\":\"On this page\"}],[[\"$\",\"a\",\"tl-dr\",{\"href\":\"#tl-dr\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"TL;DR\"}],[\"$\",\"a\",\"definition\",{\"href\":\"#definition\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"Definition\"}],[\"$\",\"a\",\"why-citation-patterns-matter\",{\"href\":\"#why-citation-patterns-matter\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"Why Citation Patterns Matter\"}],[\"$\",\"a\",\"citation-mechanics-by-engine-2026\",{\"href\":\"#citation-mechanics-by-engine-2026\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"Citation Mechanics by Engine (2026)\"}],[\"$\",\"a\",\"direct-citation-vs-synthesized-mention-vs-recommendation\",{\"href\":\"#direct-citation-vs-synthesized-mention-vs-recommendation\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"Direct Citation vs Synthesized Mention vs Recommendation\"}],[\"$\",\"a\",\"what-gets-cited-chunks-pages-and-snippet-extracted-blocks\",{\"href\":\"#what-gets-cited-chunks-pages-and-snippet-extracted-blocks\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"What Gets Cited: Chunks, Pages, and Snippet-Extracted Blocks\"}],[\"$\",\"a\",\"anchor-text-patterns\",{\"href\":\"#anchor-text-patterns\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"Anchor Text Patterns\"}],[\"$\",\"a\",\"practical-optimization-by-platform\",{\"href\":\"#practical-optimization-by-platform\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"Practical Optimization by Platform\"}],[\"$\",\"a\",\"chatgpt\",{\"href\":\"#chatgpt\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"ChatGPT\"}],[\"$\",\"a\",\"perplexity\",{\"href\":\"#perplexity\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Perplexity\"}],[\"$\",\"a\",\"google-ai-overviews-and-ai-mode\",{\"href\":\"#google-ai-overviews-and-ai-mode\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Google AI Overviews and AI Mode\"}],[\"$\",\"a\",\"gemini\",{\"href\":\"#gemini\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Gemini\"}],[\"$\",\"a\",\"microsoft-copilot\",{\"href\":\"#microsoft-copilot\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Microsoft Copilot\"}],\"$L34\",\"$L35\",\"$L36\",\"$L37\",\"$L38\",\"$L39\",\"$L3a\",\"$L3b\",\"$L3c\",\"$L3d\",\"$L3e\",\"$L3f\",\"$L40\",\"$L41\",\"$L42\"]]}]}]\n"])</script><script>self.__next_f.push([1,"34:[\"$\",\"a\",\"claude\",{\"href\":\"#claude\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Claude\"}]\n35:[\"$\",\"a\",\"examples-by-pattern\",{\"href\":\"#examples-by-pattern\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"Examples by Pattern\"}]\n36:[\"$\",\"a\",\"the-long-tail-reality\",{\"href\":\"#the-long-tail-reality\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"The Long-Tail Reality\"}]\n37:[\"$\",\"a\",\"common-misconceptions\",{\"href\":\"#common-misconceptions\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"Common Misconceptions\"}]\n38:[\"$\",\"a\",\"tracking-your-citations\",{\"href\":\"#tracking-your-citations\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"Tracking Your Citations\"}]\n39:[\"$\",\"a\",\"faq\",{\"href\":\"#faq\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":0,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"none\"},\"children\":\"FAQ\"}]\n3a:[\"$\",\"a\",\"q-which-platform-cites-sources-most-often\",{\"href\":\"#q-which-platform-cites-sources-most-often\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Q: Which platform cites sources most often?\"}]\n3b:[\"$\",\"a\",\"q-do-ai-citations-actually-drive-traffic\",{\"href\":\"#q-do-ai-citations-actually-drive-traffic\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Q: Do AI citations actually drive traffic?\"}]\n3c:[\"$\",\"a\",\"q-how-is-citation-rate-measured\",{\"href\":\"#q-how-is-citation-rate-measured\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Q: How is citation rate measured?\"}]\n3d:[\"$\",\"a\",\"q-can-you-force-an-ai-to-cite-your-page\",{\"href\":\"#q-can-you-force-an-ai-to-cite-your-page\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Q: Can you force an AI to cite your page?\"}]\n3e:[\"$\",\"a\",\"q-do-citation-patterns-differ-in-non-english-queries\",{\"href\":\"#q-do-citation-patterns-differ-in-non-english-queries\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Q: Do citation patterns differ in non-English queries?\"}]\n3f:[\"$\",\"a\",\"q-are-citation-patterns-stable-or-do-they-change\",{\"href\":\"#q-are-citation-patterns-stable-or-do-they-change\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Q: Are citation patterns stable, or do they change?\"}]\n40:[\"$\",\"a\",\"q-do-all-citations-carry-equal-weight\",{\"href\":\"#q-do-all-citations-carry-equal-weight\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},"])</script><script>self.__next_f.push([1,"\"children\":\"Q: Do all citations carry equal weight?\"}]\n41:[\"$\",\"a\",\"q-should-i-optimize-for-citations-or-for-rankings\",{\"href\":\"#q-should-i-optimize-for-citations-or-for-rankings\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Q: Should I optimize for citations or for rankings?\"}]\n42:[\"$\",\"a\",\"q-how-often-should-this-reference-be-reviewed\",{\"href\":\"#q-how-often-should-this-reference-be-reviewed\",\"style\":{\"display\":\"block\",\"padding\":\"4px 0\",\"paddingLeft\":16,\"color\":\"var(--text-secondary)\",\"textDecoration\":\"none\",\"transition\":\"color 0.15s ease\",\"borderLeft\":\"1px solid var(--border-default)\"},\"children\":\"Q: How often should this reference be reviewed?\"}]\n"])</script><script>self.__next_f.push([1,"12:[[\"$\",\"meta\",\"0\",{\"charSet\":\"utf-8\"}],[\"$\",\"meta\",\"1\",{\"name\":\"viewport\",\"content\":\"width=device-width, initial-scale=1\"}]]\n"])</script><script>self.__next_f.push([1,"43:I[27201,[\"/_next/static/chunks/0ld4q8u-25eux.js\",\"/_next/static/chunks/0d3shmwh5_nmn.js\",\"/_next/static/chunks/0xf6s51ok42m_.js\",\"/_next/static/chunks/148t.fhegq9f1.js\",\"/_next/static/chunks/05cra..ka3fzk.js\"],\"IconMark\"]\nf:null\n"])</script><script>self.__next_f.push([1,"14:[[\"$\",\"title\",\"0\",{\"children\":\"AI Citation Patterns: How AI Engines Cite Sources (2026)\"}],[\"$\",\"meta\",\"1\",{\"name\":\"description\",\"content\":\"Reference of how ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Gemini, Microsoft Copilot, and Claude attribute sources in 2026 — with platform-specific optimization tactics.\"}],[\"$\",\"meta\",\"2\",{\"name\":\"keywords\",\"content\":\"AI citation patterns,AI source attribution,ChatGPT citation format,Perplexity citations,Google AI Overviews citations,how AI engines cite sources\"}],[\"$\",\"meta\",\"3\",{\"name\":\"robots\",\"content\":\"noindex, follow\"}],[\"$\",\"meta\",\"4\",{\"name\":\"html-lang\",\"content\":\"vi\"}],[\"$\",\"link\",\"5\",{\"rel\":\"canonical\",\"href\":\"https://geodocs.dev/reference/ai-citation-patterns\"}],[\"$\",\"link\",\"6\",{\"rel\":\"alternate\",\"hrefLang\":\"en\",\"href\":\"https://geodocs.dev/reference/ai-citation-patterns\"}],[\"$\",\"link\",\"7\",{\"rel\":\"alternate\",\"hrefLang\":\"x-default\",\"href\":\"https://geodocs.dev/reference/ai-citation-patterns\"}],[\"$\",\"meta\",\"8\",{\"property\":\"og:title\",\"content\":\"AI Citation Patterns: How AI Engines Cite Sources (2026)\"}],[\"$\",\"meta\",\"9\",{\"property\":\"og:description\",\"content\":\"Reference of how ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Gemini, Microsoft Copilot, and Claude attribute sources in 2026 — with platform-specific optimization tactics.\"}],[\"$\",\"meta\",\"10\",{\"property\":\"og:type\",\"content\":\"article\"}],[\"$\",\"meta\",\"11\",{\"property\":\"article:published_time\",\"content\":\"2025-04-25\"}],[\"$\",\"meta\",\"12\",{\"property\":\"article:modified_time\",\"content\":\"2026-05-01\"}],[\"$\",\"meta\",\"13\",{\"property\":\"article:author\",\"content\":\"Geodocs Research Team\"}],[\"$\",\"meta\",\"14\",{\"name\":\"twitter:card\",\"content\":\"summary_large_image\"}],[\"$\",\"meta\",\"15\",{\"name\":\"twitter:title\",\"content\":\"AI Citation Patterns: How AI Engines Cite Sources (2026)\"}],[\"$\",\"meta\",\"16\",{\"name\":\"twitter:description\",\"content\":\"Reference of how ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Gemini, Microsoft Copilot, and Claude attribute sources in 2026 — with platform-specific optimization tactics.\"}],[\"$\",\"link\",\"17\",{\"rel\":\"shortcut icon\",\"href\":\"/favicon.ico\"}],[\"$\",\"link\",\"18\",{\"rel\":\"icon\",\"href\":\"/favicon.ico?favicon.0zl.ysuv3a32n.ico\",\"sizes\":\"48x48\",\"type\":\"image/x-icon\"}],[\"$\",\"link\",\"19\",{\"rel\":\"icon\",\"href\":\"/favicon.ico\"}],[\"$\",\"link\",\"20\",{\"rel\":\"apple-touch-icon\",\"href\":\"/favicon.ico\"}],[\"$\",\"$L43\",\"21\",{}]]\n"])</script></body></html>