How to audit AI Overviews visibility (Google): checklist + metrics

TL;DR

Do this in five steps: (1) build a 50-200 query set covering branded, comparison, and informational intents; (2) capture each query in a logged-out, location-controlled environment and store the AIO HTML or screenshot; (3) score each result on six fields (AIO triggered, brand cited, brand mentioned, citation position, sentiment, competitor citations); (4) compute five headline metrics; (5) compare against the previous month's frozen baseline. Re-baseline only when the query set changes.

Why a separate AI Overviews audit

Google AI Overviews now appear above traditional blue links for the majority of informational queries and an increasing share of commercial ones. They generate answers from a curated set of cited sources rather than ranking pages by themselves. A traditional SEO audit measures rank, impressions, and CTR — none of which capture whether your brand is the cited source in the answer the user actually reads.

Industry tracking tools and audit playbooks (Meltwater's GenAI Lens, Averi's Brand Visibility Score, IdeaHills' AIO audit, ZipTie's AI search readiness checklist, Data-Mania's 2026 AIO checklist, and the Lily Grozeva audit framework on LinkedIn) converge on the same conclusion: AIO visibility is a separate KPI that requires its own data collection.

Audit prerequisites

Before running the checklist, lock down the following:

• [ ] Stable environment. Audits are run logged-out, in a clean browser profile or an automation runner, with a fixed locale (country + language) and a fixed device class (desktop and mobile reported separately).
• [ ] Frozen query set v1. A versioned spreadsheet of queries with intent labels. Never edit queries mid-cycle; bump the version when you change the set.
• [ ] Storage layout. One row per (query × device × date) with screenshot path, full HTML capture, and structured fields.
• [ ] Owner. A single person responsible for monthly execution. The audit fails silently if it is everyone's part-time job.

The checklist

1. Build the query set

• [ ] 50-200 queries total, depending on category breadth.
• [ ] Mix of intents:
• Branded (your brand + variants).
• Branded competitor (each top-5 competitor).
• Comparison ("X vs Y", "alternatives to X").
• Informational / how-to (top of funnel).
• Transactional / commercial ("best X for Y", "X pricing").
• Long-tail conversational phrased as a question.
• [ ] Include a small adversarial set of queries where misinformation about your brand has appeared (Reddit-driven, outdated facts, common confusions).
• [ ] Each query is tagged with intent, funnel_stage, and priority (P0-P2).
• [ ] Sign-off: stakeholders agree the set represents the buyer journey before v1 is locked.

2. Capture

• [ ] Run each query in your controlled environment. Capture both desktop and mobile.
• [ ] Save the full SERP HTML (or use a SERP API that captures AIO blocks).
• [ ] Save a full-page screenshot for visual review and stakeholder reporting.
• [ ] Record structured fields for every query in the audit log:
• aio_triggered (boolean).
• aio_word_count.
• aio_citations[] (URL, anchor text, position 1-N).
• brand_cited (boolean) and brand_mention_count (the brand name appearing in the answer body even without a citation).
• competitor_citations[] (per top-5 competitor).
• sentiment (positive / neutral / negative / inaccurate).
• aio_block_type (text-only, list, table, follow-ups).
• [ ] Time-stamp every capture; AIO output is volatile and can change within minutes.

3. Quality-control the captures

• [ ] Manually review a 10% sample for parser errors (especially URL extraction from AIO source cards).
• [ ] Flag captures where the AIO failed to load or the page errored — do not score these.
• [ ] Confirm a baseline run from the previous cycle is on disk for diff comparison.

4. Score and compute metrics

• [ ] Compute the five headline metrics (definitions in the next section).
• [ ] Compute per-slice metrics: by intent, funnel stage, device, and priority tier.
• [ ] Identify wins (queries where you newly appear or move up in citation position) and losses (queries where you drop out).
• [ ] Identify misinformation findings (negative or inaccurate sentiment captures) and route to the AI citation crisis response checklist.

5. Report

• [ ] One-pager with the five headline metrics vs prior month, plus deltas.
• [ ] Per-section coverage table (informational, comparison, transactional).
• [ ] Top-5 wins, top-5 losses with screenshots.
• [ ] Action list: queries to target with new content, pages to refresh, citations to pursue.
• [ ] Archive the full audit folder under audits/aio/ so prior cycles remain reproducible.

The five headline metrics

Keep the dashboard small. These five capture 90% of useful signal.

1. AIO trigger rate

trigger_rate = queries_with_AIO / total_queries

Measures the share of your priority queries that Google now answers with an AI Overview. A rising trigger rate means more of your funnel is answered above the blue links.

2. Citation rate

citation_rate = queries_where_brand_is_cited / queries_with_AIO

The core success metric: when an AIO triggers, how often is your brand one of the cited sources? SaaS leaders typically target 20-30% citation rate on their priority query set as the threshold of meaningful AI visibility; most starting brands measure under 5%.

3. Mention rate

mention_rate = queries_where_brand_is_mentioned / queries_with_AIO

Mentions are when the answer body says your brand name without citing your page. Mention rate captures recognition even when you didn't earn the link, and helps you identify content opportunities where your brand is recognized but other sources are getting the citation.

4. Average citation position

avg_position = mean(citation_position_when_cited)

Lower is better. Position 1-3 in an AIO source carousel earns disproportionately more click-through than positions 4+. Track distribution, not just the mean.

5. Share of AI voice

SoAV = brand_citations / (brand_citations + sum(competitor_citations))

Your share of all citations across your priority query set, normalized to your competitive set. SoAV is the single most useful number for executive reporting because it isolates competitive movement from search-volume changes.

Tools you can use

• Manual + spreadsheet. Smallest setup. Sufficient for a 50-query audit run by one person monthly.
• SERP APIs (Semrush AIO tracker, Ahrefs, Sistrix, Serpapi). Automate capture of structured fields.
• Dedicated AIO/AI search trackers (Peec.ai, Otterly, Profound, RankScale, Knowatoa, Meltwater GenAI Lens). Add cross-engine coverage (ChatGPT, Perplexity, Gemini) and time-series storage.
• Schema validators + GSC. For diagnostics on cited pages: confirm structured data, validate canonical, monitor impressions on the cited URLs.

Do not adopt a paid tool until your manual spreadsheet audit is producing month-over-month deltas your team trusts. Tooling accelerates an audit that already works; it cannot rescue one that doesn't.

Common pitfalls

• Unstable query set. Adding or removing queries between cycles destroys longitudinal comparisons.
• Logged-in captures. Personalization changes results; always run logged-out.
• Ignoring mobile. AIO behavior differs by device; report both.
• Counting raw mentions as citations. A mention without a link is recognition, not a citation. Score them as separate fields.
• No competitor baseline. Citation rate alone hides a category-wide AIO expansion that lifts everyone.
• One-shot audits. Single-run AIOs are noisy; AirOps research found only ~20% of brands stay visible across five consecutive runs of the same query. Capture each query 2-3 times in the same week and report the median.
• Audit without action. A monthly audit that doesn't feed the editorial roadmap is reporting theatre.

FAQ

Q: How often should I run the audit?

Monthly is the right cadence for most brands. Weekly is overkill given AIO volatility (you'll capture noise more than signal). Quarterly misses content-change effects entirely.

Q: How big should the query set be?

Start with 50 priority queries and grow to 150-200 as your team scales. Below 30 queries, per-slice metrics become unreliable. Above 300, manual review becomes infeasible without automation.

Q: Should I include zero-volume queries?

Yes for buyer-journey completeness, no for trigger-rate denominators. Track them in a separate "long-tail diagnostic" tab so they don't distort the headline numbers.

Q: How do I handle inaccurate AIO answers about my brand?

Flag them in the capture log with sentiment = inaccurate, then route to the AI citation crisis response checklist for the response playbook (corrections via authoritative pages, schema updates, third-party correction outreach).

Q: Do I need a paid tracking tool?

Not to start. Manual + spreadsheet covers a 50-query monthly audit in under two hours once the workflow is in place. Adopt a paid tool when (a) you scale beyond ~150 queries, (b) you need cross-engine coverage (ChatGPT/Perplexity/Gemini), or (c) you need to share dashboards with stakeholders who won't open a spreadsheet.

Q: How do I make the metrics survive Google's frequent AIO changes?

Keep the metric definitions stable while letting raw counts move. Trigger rate, citation rate, mention rate, position, and share of AI voice are all relative measures, so they remain comparable across AIO surface changes as long as the query set is frozen and the capture method is consistent.