AI Tracker

Visibility measures whether engines name you. AI Tracker measures what happens next: the humans who read an AI answer and click through to your site, and the crawlers that read your pages to build those answers.

Standard analytics cannot do this reliably. AI referrals arrive with sparse or missing referrers, are frequently misfiled as direct traffic, and are lost entirely when a script is blocked. The Geonimo tracker is first-party and purpose-built for the distinction.

Preview. The tracker ingest endpoint is not yet live, so this screen renders a sample dataset. Installing the script now does not yet produce data. See Product status.

Installing

Add one line before </head>, or through your tag manager:

<script async src="https://cdn.geonimo.com/tracker.js" data-project="YOUR_PROJECT_ID"></script>

Your project ID is on the AI Tracker screen, next to a copy button. The script is asynchronous and does not block rendering.

Verify by visiting your site and checking that a session appears under Recent sessions. Give it a minute.

What it separates

Every visit is classified into one of three segments, and the whole screen respects that split:

• AI — human visitors arriving from an AI assistant.
• Organic — search-engine referrals.
• Rest — direct, social, email, referral, everything else.

AI bot crawls are tracked separately from AI-referred humans. Both matter, for different reasons: crawls tell you the engines can read your pages, referrals tell you the answers convert to attention.

The KPI row

Sessions, Pageviews, Average duration and Average scroll depth for the current segment and date range — each with its change over the period.

Scroll depth is included because it is the honest engagement metric for AI traffic. Visitors who arrive from an assistant tend to arrive informed and read further down the page than search visitors, and average duration alone hides that.

Quality by AI platform

A card per AI source — ChatGPT, Perplexity, Google AI surfaces and others — with sessions, pageviews, average duration, average scroll and engaged sessions. Open one for the full breakdown for that platform, including its own recent sessions.

Platform quality varies enough to change strategy. Traffic from one assistant may convert twice as well as another at a fraction of the volume; that is an argument for where to spend optimization effort, and it only shows up if you measure per platform.

Segment comparison

A table comparing AI, Organic and Rest across average session duration, pages per session, engaged sessions and conversion rate, each with its trend.

This is the table for the quarterly review. The typical finding — AI traffic being lower volume but markedly higher intent — is the entire commercial argument for GEO, and it is better made with your own numbers than with a vendor's benchmark.

Breakdowns

• Top landing pages — which pages AI answers actually send people to, with AI sessions, average duration and engagement. Rarely your homepage; usually a specific comparison, guide or tool page.
• Devices, Browsers, Countries — the usual dimensions, filtered to the current segment.
• Recent sessions — individual sessions you can open to see the pages visited in order.

Session history is the qualitative layer: the path a single AI-referred visitor took through your site tells you what the answer promised and whether the page delivered it.

What to do with it

1. Confirm the classification. Check a handful of recent AI sessions and make sure they look right before building a report on them.
2. Read the landing pages. The pages AI sends traffic to deserve the most optimization attention — they are already proven to be cited. Feed them into Pages.
3. Compare quality, not volume. AI traffic is a minority of most sites' sessions today. Judge it on engagement and conversion, then decide how much of tomorrow's effort it deserves.
4. Close the loop. Rising visibility with flat AI sessions means you are being mentioned without being compelling — an answer-content problem, not a coverage problem.

Privacy

The tracker is first-party, collects pageviews, sessions and engagement signals, and is designed to work without third-party cookies. Treat it as you would any analytics script in your privacy policy and consent flow. See Security and privacy.