AI Visibility for Businesses: How to Measure It and Track It

Most of what is written about AI visibility tells you how to win it. Fewer pieces start where any serious effort actually has to start: measurement. Since AI Overviews became a standard feature of Google search and ChatGPT and Perplexity began citing their sources, "are we cited?" has turned from a vague worry into a number you can put on a dashboard. And a number you can put on a dashboard is a number you can move. This guide focuses on that first, often-skipped step, so the work that follows aims at something real.

What "measuring AI visibility" actually means

Measuring a business's AI visibility means putting a panel of real business questions to ChatGPT, Google AI Overviews and Perplexity, then recording, query by query, whether your company is cited as a source or named in the generated answer. The share of queries where you appear is your citation rate.

The discipline of optimising for these engines has a name, GEO, for Generative Engine Optimization, formalised in 2023 by researchers from Princeton, Georgia Tech, the Allen Institute and IIT Delhi in a paper presented at the ACM SIGKDD conference. Their central insight is what makes measurement possible: a generative engine does not return a page, it synthesises an answer from several sources and chooses which ones to cite according to criteria you can observe. If the choice is observable, the outcome is measurable.

So measurement is not a vibe check. It is a structured reading of who gets named when a prospect asks a question. The unit is the query, not the keyword. Where classic SEO asks "where do I rank for this term?", AI visibility measurement asks "when someone asks this question, does the answer mention me?" Those are different questions, and they deserve their own instrument.

Why measurement comes before optimisation

Without a baseline, you cannot tell whether your content work is moving the needle or wasting effort. Measurement first gives you a starting point, a list of the exact queries you are losing, and the competitors cited in your place, which is the raw material for everything that follows.

It is tempting to skip straight to producing content. We see it constantly. A business hears that AI visibility matters, commissions a stack of articles, and a few months later has no idea whether any of it worked, because nobody wrote down where things stood before. The number you did not take on day one is the number you can never recover.

Measurement also reframes the work as a list of concrete targets. When you run a query panel and find your business absent from fifteen commercial-intent questions, you do not have a vague problem, you have fifteen briefs. Each missing answer is a query to win, and the competitor cited in your place tells you the bar to clear. Across the dozens of businesses whose AI visibility we have checked query by query, the ones that improved fastest are almost always the ones that started from a clear, written baseline rather than a hunch.

The five-step measurement method

Measuring AI visibility follows a repeatable order: build a query panel, run each query across the engines, score the result into a citation rate, set a monthly cadence, then tie each missed query to a content action.

This is the exact sequence we apply at Cicero Studio, and one you can run yourself with a spreadsheet. The discipline is in keeping the panel stable so the numbers stay comparable from one month to the next.

1. Build a query panel of 20 to 40 questions. Write the questions a real prospect would type into an AI before choosing a supplier like you. Mix commercial intent ("best X software for a 50-person team", "who does Y in Lyon") with informational intent ("how does Z work"). Keep them phrased as a human would, not as keyword fragments.
2. Run each query across the engines. Ask each question to ChatGPT and Perplexity, and trigger Google AI Overviews on the same wording. For every query, note whether your business is cited as a source, named in the prose, both, or absent. Do it from a clean session so a personalised history does not skew the answer.
3. Score the result. Count the queries where you appear, divide by the total, and you have a citation rate per engine. Alongside it, log which competitors are cited in your place. That second column is often more useful than your own score.
4. Set a monthly cadence. Re-run the exact same panel every month. The absolute number on day one matters far less than the slope across three or four months. A stable panel is what lets you read a trend instead of noise.
5. Tie each gap to an action. Turn every missed commercial query into a content brief. After the content ships and is absorbed, watch that query's status flip from absent to cited. That is the loop that closes measurement back into results.

Measurement is a quiet, repeatable habit, not a one-off panic check before a board meeting.

The three numbers worth tracking

For a clear read, track three things: your citation rate (share of queries where you appear), your share of voice against named competitors, and your coverage of commercial-intent queries specifically. Together they tell you not just how visible you are, but whether you are visible where it pays.

It is easy to drown in metrics. After auditing many businesses, we have found that three numbers carry almost all the signal, and the rest is mostly decoration. Here is how they differ and what each one is good for.

The third one deserves emphasis. A business can post a flattering overall citation rate while being absent from every query a buyer actually asks. That is why the founding GEO study insisted on the quality of the content behind a citation, and why Google's own documentation on its AI features ties visibility to useful, reliable, people-first content rather than to a hidden toggle. Visibility on the wrong queries is a number that feels good and does nothing.

Turning a snapshot into a monthly trend

A single measurement is a snapshot; the value comes from repeating it. Re-run the identical query panel every month so the numbers stay comparable, log the result in one place, and read the slope. Monthly is the right cadence for most businesses, since AI answers shift gradually, not day to day.

The first time you measure, you get a photograph. Useful, but static. The instrument only becomes powerful when you take the same photograph from the same angle every month, because then the change between photographs is the information. A citation rate of, say, three queries out of forty tells you little on its own. The same panel reaching eight out of forty three months later, while a rival drops, tells you the content work is landing.

Why monthly and not weekly? Because AI engines re-index and absorb new content over weeks, not hours. The signals that lift content into a generated answer, as the GEO research showed, are the same patient signals that lift it in classic search: authority, freshness, structure, named sources. Daily tracking mostly captures the engines' own jitter and tempts you into reacting to noise. A calm monthly re-run respects the real rhythm of the medium. After a major content push, one extra check a few weeks later is worth it, just to confirm absorption.

This measure-then-act loop is the heart of how we work: it is the GEO audit + editorial production + automated semantic internal linking applied as a continuous cycle rather than a one-off project. For the FR-speaking version of the broader approach, our pillar on what a agence GEO does lays out each criterion we assess, and our deeper dive on the GEO audit breaks the diagnostic down step by step.

Measurement mistakes that mislead you

The traps are subtle: measuring once and calling it tracking, mixing intents so a vanity rate hides the commercial gap, letting a personalised session skew the answers, and changing the panel between runs so the numbers stop being comparable.

A measurement you trust wrongly is worse than no measurement, because it sends you confidently in the wrong direction. These are the slip-ups we see most often.

• Measuring once and stopping. A single reading is a snapshot, not a trend. Without a monthly re-run you cannot tell whether anything you did mattered. Tracking is the point, not the first photo.
• Mixing commercial and informational queries into one figure. A high overall rate can hide total absence on the queries that bring buyers. Always isolate commercial coverage as its own number.
• Querying from a logged-in, personalised session. Your own history nudges the answer. Measure from a clean session so you see what a fresh prospect sees, not a tailored echo.
• Changing the query panel between runs. Swap the questions and you lose comparability. Keep the panel fixed; add new queries as a separate, clearly dated cohort rather than editing the original.
• Reading a single competitor's presence as defeat. One rival cited once is noise. The signal is the pattern across the whole panel over several months.

The thread through all of these is the same: rigour in the measurement protects the rigour of everything you build on top of it. A loose number breeds loose decisions.

Alexis Dollé

CEO & Founder of Cicero Studio

A growth and SEO content strategy specialist, founder of Cicero Studio, I launched the agency to help businesses capture durable organic visibility, on Google as in AI answers. Every piece of content we produce is built to convert, not just to exist.

LinkedIn →

What this measurement does not tell you

For honesty, and because that transparency is exactly what AIs reward, here is what a citation measurement cannot do for you.

Measurement is the compass, not the journey. It points you at the right queries and confirms when the work is landing, but the editorial production in between is what actually earns the citations. For a business with no substantive content yet, the honest first step is to create something worth citing, then measure.

Going further

We document our method in the open, because it is our best proof. This guide handles measurement; the resources below take the neighbouring subjects one at a time, from the strategy of winning citations to the engine-specific tactics and the broader market picture. Pick the ones that match where your business is in the work: