AI Search Reporting: Dashboard Setup

TL;DR

A useful AI search dashboard answers four questions every week: how often are we cited, where are we cited, is AI sending us traffic, and is our content ready to be cited next time? Build it in three layers — a tracking layer (manual prompt logs and a dedicated AI visibility tool), an analytics layer (GA4 with UTM tagging plus Search Console), and a content readiness layer (schema, freshness, and internal-link coverage). For broader measurement context, see the Strategy hub and AI Visibility Measurement.

Pick the metrics first

A dashboard is only useful if it answers a question your team has already agreed to ask. Before wiring up tools, fix the metric set:

• Citation share. The percentage of tracked prompts where your domain or brand is cited. This is the closest analogue to rank in classical SEO. Pair with AI Search KPIs.
• Mention frequency. How often your brand is named without a linked citation. Mentions still build authority and influence future selection.
• AI referral traffic. Sessions arriving from AI engines. Requires UTM tagging on outbound citations or referrer parsing in GA4.
• Content readiness. The share of priority pages that meet the GEO content checklist (schema, answer-first format, hub linkage, freshness window).
• Competitive share of AI voice. Your citation share relative to a fixed set of competitors. Connects to the AI Search Competitive Analysis Framework.

If you cannot get a metric automatically yet, log it manually in a spreadsheet. A dashboard with five honest metrics beats a dashboard with twenty unreliable ones.

Three-layer architecture

Think of the dashboard as three layers stacked on the same prompt set.

Layer 1 — Tracking layer (what AI sees)

This layer answers "are we cited?" It needs a stable prompt library that mirrors real user queries.

• Manual prompt testing. A weekly run across ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews against 25-50 priority prompts. Log presence, position, and whether the citation links back.
• Dedicated AI visibility tools. Categories include continuous AI prompt monitors (for example Otterly, Profound, Scrunch AI, Evertune), citation features inside SEO suites (Semrush AI Toolkit, Ahrefs Brand Radar), and enterprise visibility platforms (BrightEdge, Conductor). See the Citation Monitoring Stack Selection Framework for build-vs-buy guidance. Starter tiers begin in the low double digits per month; enterprise plans are custom.
• Source attribution. Distinguish mention (brand named) from citation (brand named with a linked source). They lead to different optimization plays.

Layer 2 — Analytics layer (what users do)

This layer answers "is AI sending us traffic and what do they do?"

• GA4 referral grouping. Build a custom channel group for AI engines using referrer hostnames (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai). GA4 has no native AI source group as of 2026, so this requires manual setup or UTM tagging on outbound citations you control.
• UTM-tagged citations. When you ship content with outbound calls-to-action that AI may surface, tag the URLs with a consistent utm_source=ai-engine&utm_medium=ai-citation convention.
• Search Console for AI Overviews. AI Overviews still appear in standard SERPs and surface in Search Console performance reports. Use it as the secondary signal layer per Google Search Console for GEO Monitoring.

Layer 3 — Readiness layer (is our content ready to be picked next time?)

This layer answers "if AI looked again tomorrow, would we be picked?"

• Schema coverage. Percentage of priority URLs with valid JSON-LD; cross-reference Structured Data for AI Search.
• Answer-first format. Percentage of priority URLs with a labeled TL;DR or AI summary block.
• Freshness. Share of pages updated within the review cycle window (default 90 days).
• Internal link coverage. Number of priority pages reachable from the section hub within two clicks.

Reporting cadence

Weekly report (operations cadence)

Week of [Date]

Tracking

• Citation share (overall): X%
• Citation share by platform: ChatGPT X% | Perplexity X% | AIO X% | Claude X% | Gemini X%
• Mentions without citation: X
• New citations gained: X | Lost citations: X
• Top 5 cited URLs: [list]

Analytics

• AI referral sessions: X
• AI referral conversions: X
• Top AI-driven pages: [list]

Readiness

• Pages shipped or refreshed: X
• Schema added: X URLs
• Open issues: [list]

Decisions for next week

• [ ] ...

Monthly report (strategy cadence)

Add to the weekly:

• Share-of-AI-voice trend vs. tracked competitors.
• Content gap assessment versus competitor citation patterns.
• Topic clusters gaining or losing share.
• Tool spend and ROI snapshot.

Quarterly report (executive cadence)

A two-page brief for leadership:

• Citation share trend with annotations of major site or content changes.
• AI referral traffic vs. organic search and direct.
• Pipeline or revenue attributed (if attribution model is in place — see AI Search Attribution Model).
• Competitive movement and one strategic recommendation.

Setting up tracking, step by step

1. Build the prompt library. 25-50 prompts that span informational, comparison, and commercial intent in your category. Version it.
2. Choose a tool tier. Start with a manual prompt log + GA4. Add a dedicated AI visibility tool once the prompt library is stable.
3. Wire GA4. Create the AI engine channel group; set conversion events; document the UTM convention.
4. Define ownership. Assign one owner per layer (tracking, analytics, readiness) and one report owner.
5. Pick the dashboard surface. Looker Studio, GA4 Explorations, or whatever tool the team already opens daily. Avoid building a new home no one visits.
6. Set a cadence and stick to it. Weekly operations review, monthly strategy review, quarterly executive review.

Common pitfalls

• Tracking too many prompts. Beyond ~50 prompts, signal degrades and noise grows. Rotate prompts quarterly instead of expanding indefinitely.
• Equating mentions with citations. Mentions matter, but a linked citation is what actually drives traffic. Keep them separate in the dashboard.
• Skipping competitor benchmarks. Citation share without a denominator is hard to act on. Always pair your share with a tracked competitor set.
• Reporting once and never refining. AI engines retrain and reweight frequently. Treat the prompt library as living, not static.
• Mistaking one good week for a trend. Use a four-week rolling average for citation share before declaring a win or loss.

Stakeholder map

FAQ

Q: What is the minimum viable AI search reporting dashboard?

A prompt library of 25 priority queries, a weekly manual run across ChatGPT, Perplexity, and Google AI Overviews, plus a GA4 referral channel group for AI engines. Everything else is optimization.

Q: Do I need a dedicated AI visibility tool?

Not on day one. Start with manual logs and GA4 to validate which metrics actually drive decisions. Add a dedicated tool when the manual workload exceeds a few hours per week or when leadership wants competitor benchmarks.

Q: How often should I update the prompt library?

Review quarterly. Retire prompts that no longer reflect user intent and add prompts that emerge from search query reports, sales conversations, and customer support tickets.

Q: Can GA4 tell me which AI engine sent the visit?

Only if you build a custom channel group keyed on referrer hostname or you tag outbound citations with UTMs. There is no native AI engine grouping in GA4 as of 2026.

Q: How do I report AI search ROI to executives?

Pair citation share with downstream attribution: AI referral sessions, conversions, and revenue when an attribution model is in place. Use the quarterly cadence so trend lines are credible.