How to Measure AI Visibility

Juan Camilo Auriti · October 7, 2026

Measuring AI visibility means tracking four separate signals — a readiness score, citation presence, share of voice against named competitors, and AI-referral traffic — because no single free tool reports all four at once. This page assumes you already know what AI visibility is and want the practical "how do I actually check this for my own site" answer, built from tools that exist today, with no invented benchmarks.

The four measurable dimensions

• Readiness score — a structural measure of whether your site can be retrieved and cited: crawler access, schema depth, factual density, entity clarity. This is a proxy for retrievability and citability, checked before anyone has actually asked an AI system anything.
• Citation presence — whether an AI engine actually cited your domain for a specific prompt, right now. This is the closest thing to a direct measurement of "presence in answers," but it's a snapshot, not a trend, unless you repeat it.
• Share of voice — how often you're the cited domain versus how often a named competitor is, across the same set of prompts. It's a comparative read, not a standalone number, and it takes more than one check to mean anything.
• AI-referral traffic — actual visits arriving at your site from an AI system's answer (a link a user clicked out of ChatGPT, Perplexity, or similar). It's the only one of the four that measures downstream behavior instead of retrieval or citation.

None of these four numbers substitute for each other. A high readiness score with zero citation presence tells you the structural work is done but isn't converting into mentions yet. Citation presence with no referral traffic tells you AI systems mention you but readers aren't clicking through — which may be fine, since many AI-answer mentions are brand-visibility events with no click at all.

The DIY stack for measuring AI visibility

1. Readiness score — free GEO audit

Start with a free GEO audit: no account, no card, a full 8-category readiness report on any URL (four of the category breakdowns unlock when you leave an email address). This gives you the readiness-score dimension directly, and it's the only one of the four you can get in one pass with zero setup. Re-running it after a change is the simplest way to see whether a fix actually moved the structural signals that make citation possible in the first place.

Run the free AI SEO audit

2. Citation presence — free citation checker

To check whether an AI engine is actually citing you today, rather than just structurally ready to be cited, run a citation check against your own domain for a handful of real buyer questions. It queries Perplexity specifically — it does not query ChatGPT directly, since that engine doesn't expose sources the same way — and its sources are verifiable. Treat it as one engine's view, and since the same query can return different sources run to run, as a sample rather than a verdict. One check is a snapshot; run the same prompts again weeks later to see whether presence is trending up, down, or flat.

3. Share of voice — sample manually, then automate what you can

Share of voice needs the same fixed prompt set asked about a named competitor, not just yourself. You can sample it manually: ask the same handful of questions in a clean AI session and tally which domains get cited each time. On Studio, automated citation tracking (5 Perplexity checks a day) saves the answers for a tracked prompt set, and the share-of-voice view reads that saved history to show whether a named competitor is mentioned more often than you — on Perplexity only, not ChatGPT or Google's AI Overviews. For the fuller workflow on picking a competitor set and reading the resulting deltas, see how to benchmark your AI visibility against competitors — this page names share of voice as a dimension; that page is where the step-by-step method lives.

Worth being precise about one adjacent feature: Studio's "competitor comparison" line item compares structural GEO-score readiness between domains, not citation counts. It's a useful, real feature — just a different dimension than share of voice, and this page doesn't conflate the two.

4. AI-referral traffic — segment it in GA4

If you already have GA4 on your site, you can isolate AI-referred sessions without any new tool:

• Open Reports → Acquisition → Traffic acquisition (or build an Exploration for more control), and add a filter or comparison on Session source / medium.
• Filter for the AI referrers relevant to your audience — common ones include chatgpt.com / referral, perplexity.ai / referral, gemini.google.com / referral, and copilot.microsoft.com / referral. Exact source strings vary by GA4 setup and change as AI products update their referrer behavior, so check what actually shows up in your own data rather than assuming a fixed list.
• Save it as a segment or a comparison so you can watch the trend over successive weeks, the same way you'd watch organic-search traffic, rather than reading one week in isolation.
• Treat a rising trend as directional, not as a benchmark against anyone else's numbers — GA4 referral data is inherently about your own site's traffic mix, and there's no published industry baseline to compare it against honestly.

This is a qualitative, "how to look" method, not a tool with its own dashboard — GA4 is doing the segmentation, GeoReady isn't a traffic-analytics product.

One honest limitation worth stating plainly: not every AI-answer click arrives with a usable referrer. Some AI apps pass the referring domain cleanly; others strip it, or route the click through an intermediate redirect that GA4 logs as (direct) / (none) instead of the AI product's domain. That means AI-referral segmentation will typically undercount real AI-driven traffic, not overcount it — a rising trend in the segment you can isolate is a reliable directional signal, but a flat or low number doesn't necessarily mean AI systems aren't sending you readers, only that some of them arrive looking like direct traffic.

Why no single tool covers all four

Part of why this is a stack of methods rather than one report is that the four dimensions answer genuinely different questions, aimed at different audiences: a readiness score is about your own site's structure, independent of any specific question a buyer might ask; a citation check is about one specific prompt, right now; share of voice only means anything relative to a named competitor; and referral traffic is the only dimension that reflects actual reader behavior rather than a model's retrieval or citation decision. A vendor claiming to collapse all four into a single dashboard number would be hiding which of these four very different questions it's actually answering.

How often to repeat all of this

A one-time measurement across these four dimensions tells you where you stand today; it doesn't tell you whether that's changing. How frequently to re-run the readiness score, the citation check, and the referral-traffic review is a separate question from what to measure — covered in full in how often to monitor AI visibility, which frames cadence by how fast a site changes rather than "as often as possible."

Start measuring

Run the free GEO audit first — it's the fastest way to get the readiness-score dimension with no setup. Layer in a citation check next, then GA4 segmentation, once you have a baseline worth tracking.

Run the free AI SEO audit