Generative Engine Optimisation Reporting in GA4: A Dashboard Spec for 2026

How do I report on GEO performance in GA4?

A useful GEO (Generative Engine Optimisation) dashboard combines three data sources in one Looker Studio report: GA4 (for AI-channel sessions, engagement, and conversions), Search Console (for AI-likely query CTR via the heuristic filter approach, since there's no native AIO filter), and a third-party citation tracker (Profound, Otterly.AI, Peec AI, or manual sampling).

Six widgets are essential: AI traffic by source, engagement vs organic baseline, conversion rate by AI source, top AI landing pages, citation-frequency overlay, and month-over-month growth trend. Build time: 90 minutes for the basic version, 1 day for the full version with citation overlay.

This spec gives you the exact dashboard structure, the data joins, and the citation-tracking workflow we use across audited properties.

The two metrics you cannot avoid

Before you build anything, internalise the two metrics that define GEO performance — they're orthogonal and you need both:

Citation frequency. How often your domain appears as a source in an AI-generated answer. Measured *outside* GA4, because AI engines mostly don't tell you you've been cited unless someone clicks. Sources: Profound, Otterly.AI, Peec AI, AlsoAsked, or manual weekly sampling of 50 priority queries across ChatGPT, Perplexity, Gemini, Claude, Copilot.

Click-through rate from citations. The percentage of citations that produced a click to your site. Measurable in GA4 via your AI-traffic custom channel, divided by the impression data from your citation-tracking tool.

A page can have high citation frequency and zero clicks (typical for AI Overview citations — answers satisfy intent on the SERP). A page can have low citation frequency but high CTR per citation (typical for high-intent commercial queries where users specifically want the source). Both states require different responses. The dashboard has to surface both.

The six essential widgets

Widget 1 — AI traffic by source (GA4 custom channel group)

Data source: GA4

Visualisation: Time-series chart, stacked area

Dimension: Session source (filtered to AI sources via custom channel group)

Metric: Sessions

Date range: Trailing 90 days

This is your headline metric — the trend of AI-driven sessions over time, stacked by platform (ChatGPT, Perplexity, Gemini, Copilot, Claude, etc). The stack tells stakeholders which AI surfaces are growing for your business; the total tells them how big the AI channel is overall.

To enable: build the AI Assistants custom channel group in GA4 first (see the *ChatGPT, Atlas, Perplexity, Comet, Claude* post for the regex). Then in Looker Studio, use Session source as the dimension with this regex filter:

Widget 2 — Engagement vs organic-search baseline

Data source: GA4

Visualisation: Bar chart, side-by-side

Dimension: Session custom channel group (AI Assistants vs Organic Search)

Metrics: Engagement rate, Average engagement time per session

Date range: Trailing 90 days

This widget answers the "is AI traffic actually any good?" question every CMO asks. AI visitors typically engage 23% more than non-AI traffic and stay 41% longer (Adobe, Q2 2025). If your AI engagement rate underperforms organic, your custom channel group regex is catching false positives (most likely culprit: a too-loose pattern matching email or other non-AI sources). Refine the regex.

Widget 3 — Conversion rate by AI source

Data source: GA4

Visualisation: Table sorted by conversion rate descending

Dimensions: Session source/medium

Metrics: Sessions, Key event count, Key event rate

Filter: Session custom channel group = AI Assistants

Date range: Trailing 90 days

This shows which AI sources actually convert. AI traffic averages 15–27% conversion rate vs 1.76% for Google organic (Seer Interactive 2025; Broworks 2026). Within AI traffic, Perplexity tends to outperform ChatGPT for B2B; Claude tends to outperform ChatGPT for technical content. Use this widget to inform where to invest GEO effort.

Widget 4 — Top AI landing pages

Data source: GA4

Visualisation: Table sorted by sessions descending

Dimensions: Landing page + query string

Metrics: Sessions, Engagement rate, Conversions

Filter: Session custom channel group = AI Assistants

Date range: Trailing 90 days

This is the most operationally valuable widget. The pages here are your highest-cited content — the pages AI assistants are actually quoting and sending traffic to. Three uses:

1. Refresh priority list. Top 10 pages here are where monthly refreshes pay back fastest.
2. Content cluster expansion. Topics with multiple pages in the top 20 are clusters where you have momentum and can build out more spokes.
3. Reverse-engineering signal. Look at what these pages have in common — answer-first structure, specific data points, FAQ schema. The patterns are your GEO recipe.

Widget 5 — Citation frequency overlay

Data source: Third-party tool (Profound / Otterly.AI / Peec AI) blended with GA4

Visualisation: Combo chart — bar (citations) + line (sessions) on dual y-axis

Dimensions: Date (weekly granularity)

Metrics: Citation count (from tool), Sessions (from GA4)

Date range: Trailing 90 days

This is the killer widget — it visualises the relationship between citations (impressions) and sessions (clicks). When citation count rises but sessions don't, AI engines are quoting you without sending traffic (Great Decoupling pattern). When sessions rise without citation increase, you're getting better CTR per citation — your content is becoming more click-worthy.

To build without a paid tool: do manual weekly citation tracking on 50 priority queries and store in a Google Sheet. Connect the sheet as a Looker Studio data source. The setup takes 30 minutes per week to maintain — worthwhile until you have 100+ queries to monitor.

Widget 6 — Month-over-month growth trend

Data source: GA4

Visualisation: Scorecard with comparison to previous period

Metrics: AI sessions, AI conversions, AI revenue (if e-commerce)

Filter: Session custom channel group = AI Assistants

Date range: Last 30 days vs prior 30 days

The boardroom number. AI referral traffic grew 50%+ year-over-year on most B2C sites in late 2025; Gemini referral traffic grew 388% YoY (Similarweb / Digiday, late 2025). Show these against your own MoM number to put performance in industry context.

How to handle the GSC data layer

GSC presents a problem because — as covered in *Tracking AI Overview Impressions in Google Search Console* — there is no native AI Overview filter. AI Mode, AI Overview, and standard organic data are blended under "Web". The May 2025–April 2026 impression-inflation bug compounds the issue.

The pragmatic approach for GEO dashboarding:

Don't put GSC AIO impression data in your headline metrics. It's too unreliable. Use clicks (which Google confirmed were not affected by the bug).

Do include a heuristic AI-likely-query CTR widget as a contextual metric. Build a Looker Studio chart on GSC data with a regex filter on Query matching:

This catches conversational, AI-eligible queries. Show the CTR for this segment vs your overall property CTR. The gap is your "AI Tax" — the percentage of click-through-rate you're losing to AI Overview cannibalisation. Most sites we audit show 25–55% AI Tax on informational queries.

Annotate the four discontinuity dates on every Looker Studio chart with a reference line: 13 May 2025, 17 June 2025, 12 September 2025, 3 April 2026. Add a note explaining each. This protects every YoY comparison from misinterpretation.

Citation tracking: paid vs manual

The citation-frequency widget needs source data. Three approaches:

Manual sampling (free, ~30 min/week). Pick 50 priority queries. Each Monday, query each in ChatGPT, Perplexity, Claude, Gemini, Copilot. Record in a Google Sheet: query, week, platform, cited (yes/no), citation position (if numbered). Connect the sheet to Looker Studio. Best for businesses just starting GEO measurement, or to validate paid tools.

Profound (~£500–£2,000/mo). Most comprehensive — covers ChatGPT, Perplexity, Gemini, Claude, AIO. Tracks brand mentions and citation changes. Best for large portfolios or agencies running multiple clients.

Otterly.AI (~£50–£500/mo). Cheaper alternative covering the major engines. Best for individual sites or small agencies.

Peec AI (~£200–£800/mo). Strong on AIO-specific tracking and competitive comparisons.

Semrush AI Toolkit (£100–£250/mo on top of Semrush). Bolt-on if you already use Semrush for traditional SEO — handy for combining keyword research with AI visibility.

The decision rule: under 100 queries to monitor, manual sampling is fine. 100–500 queries, use Otterly. 500+ queries or multi-client agency work, Profound or Peec.

What to put on the front page (one screen, no scrolling)

Most stakeholders only look at the first screen. Put the four most important widgets above the fold:

1. AI traffic by source — time series (top left, large)
2. MoM growth scorecard (top right, prominent)
3. Engagement vs organic baseline — bars (middle left)
4. Conversion rate by AI source — table top 5 (middle right)

The deeper widgets — top landing pages, citation frequency overlay, AI Tax CTR — go on a second tab or below the fold. They're for analysts to drill into, not for executives to scan.

Refresh cadence and review rituals

A GEO dashboard is only useful if someone reviews it. The cadence we recommend:

Weekly (15 minutes, owned by analyst). Check AI sessions for unusual movements. Review the citation-tracking sheet for queries gained or lost. Note anything for the monthly review.

Monthly (60 minutes, owned by analyst, presented to marketing director). Full dashboard walkthrough. Identify the top 3 cited pages this month and the bottom 3 cited pages where you expected presence. Build a refresh list of 5–10 pages for the next 30 days.

Quarterly (3 hours, full marketing team). Update the AI custom channel group regex (new platforms launch monthly). Review the AI Tax trend. Re-prioritise GEO investment based on which AI sources are growing for your business specifically. Renew or change citation-tracking tooling.

What this dashboard does NOT measure

Set expectations honestly. The dashboard cannot tell you:

• AI traffic from mobile apps (ChatGPT iOS/Android, Meta AI in WhatsApp, Perplexity mobile). Referrer is stripped — these arrive as Direct.
• AI traffic from ChatGPT Atlas browser. Referrer stripped, browser reports as Chrome 141 — indistinguishable from standalone Chrome.
• AI Overview citations where users didn't click. You'll see impression data via heuristic methods but never a precise count.
• Brand mentions in AI answers without a citation link. Some AI engines mention brands without linking — invisible to all dashboards.

Industry estimates put true AI traffic at 2–3x what analytics platforms report (multiple 2025–26 studies). Frame the dashboard's numbers as a floor, not a ceiling.