AI search visibility: how to audit your brand in ChatGPT and Google

AI search visibility is your brand's place in the answers of AI assistants and answer engines: ChatGPT, Gemini, Perplexity, Claude, and now Google's AI Overviews and AI Mode. Auditing it means answering three questions: is your brand cited, are your competitors cited instead of you, and which sources do the engines use to answer. The method relies on a panel of questions asked repeatedly, not on a single tool.

It is the audit I have done most often over the past year, and almost always for the same reason: a business owner asked ChatGPT to recommend a provider in their industry, didn't show up, and wants to understand why.

Start with the simplest part: your question panel. Enter your business, and the tool writes the questions your customers actually ask AI tools, ready to copy.

What is AI search visibility?

It is what SEO has measured for twenty years, carried over to answer engines. In SEO, you track your rankings on a list of keywords. In AI search visibility, you track your citations on a list of questions.

Three things make it different. First, there is no stable ranking: either you are cited or you are not, and the order changes from one answer to the next. Second, the answer varies from one session to another, which means you have to repeat the measurement. Third, no conversational AI vendor gives brands a dashboard: there is no Search Console for ChatGPT.

Google has been the exception since this summer, for its own answers. The generative AI report in Search Console, open to all sites since August 31, 2026, counts your impressions in AI Overviews and AI Mode. It is valuable data, but partial: it tells you how many times a page appears, not on which questions or against which competitors.

That is why an AI visibility audit is a method first, and only then a tool.

How to audit your AI search visibility in 5 steps

Step 1: build the question panel

This step determines the quality of everything else. The panel must reflect what your customers actually ask, not what you would like them to ask. I always build four families of questions, in roughly equal proportions.

• Recommendation questions. "Which provider for this thing in this place." These matter most commercially.
• Comparison questions. "What is the difference between this solution and that one." They reveal who the engines see as your competitors.
• Expertise questions. The technical problems of your trade. These build the brand's authority in the medium term.
• Brand questions. "Is this company any good." They show what the engine knows about you, and sometimes what it makes up.

Forty questions strike a good balance between representativeness and workload. Below twenty, the results don't mean much.

Step 2: ask the questions, properly

The rigor of the protocol is what gives the measurement its value. Three rules, and they are not negotiable.

1. Always logged out. A private browsing window, no account signed in, no history. Otherwise you measure what the engine has learned about you, not what it tells a stranger.
2. Always the same wording. A rephrased question gets a different answer: you would no longer be comparing anything from one week to the next.
3. Always several times. Two or three runs per question, on different days. A single answer is not data, it is an anecdote.

For a complete ChatGPT visibility audit, add at least Gemini, Perplexity and Google. Since July 22, 2026, AI Overviews and Google AI Mode have been live in France: they are now the AI answers your customers see most, since they show up in the search engine they already use. For each answer, I note only three things: is the brand cited, which competitors are cited, and which sources are linked.

Step 3: calculate your share of voice

Share of voice is the ratio between the number of times you are cited and the total number of brand citations across your panel. It is the metric that lets you track progress against the market. The four audit metrics are detailed just below.

Step 4: analyze the competing sources

This is the most profitable step, and the most neglected. When the engine doesn't cite you, it cites someone. List those sources: directories, comparison sites, blog posts, forums, institutional websites.

That list becomes your presence action plan. If three comparison sites in your industry keep coming back, being listed on them is worth more than ten articles on your own blog.

Step 5: turn it into a content plan

The questions in your panel that you answer nowhere are, quite literally, your editorial calendar. It is the simplest and most reliable way to choose your topics: they come from real requests, not from a search volume tool.

Measuring brand visibility in AI search: the 4 metrics

To make it concrete, a fictional example: out of 40 questions asked, your brand appears in 6 answers. Your citation rate is 15%. If those 40 answers cite 50 brands in total, 6 of them yours, your share of voice is 12%. Track these two numbers as absolute values from one measurement to the next: "6 questions out of 40" says more than a growth percentage.

The last metric often surprises my clients. It regularly happens that an engine attributes to a company a business it isn't in, a city where it isn't located, or made-up prices. Fixing that comes before everything else, because false information spreads faster than an absence.

Free ChatGPT visibility check: the half-day version

You can run a serious first audit without spending a cent. Here is the condensed version.

1. Generate twenty questions with the tool above, or write them by hand.
2. Open a private browsing window and ask them to ChatGPT, then to Perplexity, Gemini and Google in AI Mode.
3. Record in a spreadsheet: question, engine, brand cited yes or no, competitors cited, sources linked.
4. Count. Your citation rate is the number of yeses divided by the number of questions.
5. Do it again in a month, after your first fixes.

A one-off measurement is free; repetition is what turns it into an audit. What money buys here is time and automation, not data you couldn't get yourself.

AI search visibility tools: which to choose, and at what price

Beyond a few dozen questions tracked every month across several engines, the spreadsheet becomes a chore. AI visibility tracking tools ask your questions (prompts) automatically, every day, and calculate the metrics. Here are the ones I recommend looking at, with their public price checked on October 7, 2026 on each vendor's website (these price lists change often).

Google Search Console: the generative AI report (free)

Open this before anything else. The generative AI performance report shows, page by page, how many times your URLs appear in AI Overviews and AI Mode, by country, device and date (a separate report covers Discover). It does not give clicks yet, or the questions asked. It is the only official AI visibility figure available, but it only covers Google.

Free checkers: a snapshot, not tracking

Ahrefs and Semrush each offer a free AI visibility checker, and Amplitude a free report on what AI says about a brand. They give you an order of magnitude in two minutes, which helps you decide whether the topic deserves an audit. They do not replace a panel built on your own industry questions.

Otterly.AI: the most accessible

The Lite plan costs $29 a month ($25 billed annually) for 15 prompts tracked daily on ChatGPT, AI Overviews, Perplexity and Copilot. AI Mode, Gemini and Claude are available as add-ons. The Standard plan goes up to $189 a month for 100 prompts. It is a good starting point to track the ten or fifteen recommendation questions that matter commercially.

Qwairy: lots of questions for the price

The Starter plan is listed at €79 excl. VAT a month (€65 billed annually), for up to 100 prompts, with the main AI engines and more than 45 languages. Higher plans add multiple workspaces, an API and fact-checking. Good value for volume for a mid-sized business that wants to track a full panel.

Peec AI: built for marketing teams

The Starter plan costs €85 a month for 50 prompts tracked daily on three models of your choice, including ChatGPT, AI Mode, AI Overviews, Gemini and Copilot. Peec AI goes deep into analyzing cited sources and competitors. The Pro and Advanced plans go up to 150 and 350 prompts.

Semrush: if you are already on it

The SEO plan at €134 a month already includes AI visibility reports. Daily tracking of 50 prompts requires the Starter SEO + AI plan, at €190 a month (€158.40 billed annually). The AI Visibility Toolkit is also available as an add-on at $99 a month, according to Semrush's documentation. The benefit: combining AI visibility and SEO data in the same tool.

Ahrefs Brand Radar: the market view

Brand Radar is sold on its own. Tracking your own prompts starts at $50 a month, and the AI visibility index, which covers hundreds of millions of prompts modeled from Ahrefs' keyword database, starts at $199 a month. Ahrefs' paid subscriptions include a few tracked prompts. It is the most complete tool for comparing a brand to its market, especially if the brand is already well known.

AI visibility tracking: how often to measure

A full audit twice a year is enough for most businesses. In between, light tracking of ten priority questions once a month detects movements without tying up a full day. That is where a tool costing a few dozen euros a month pays for itself.

Two events, however, call for an immediate measurement: a new model release from one of the major players, which often reshuffles the deck within a few days, and a redesign of your website, which can make pages the engines used to cite disappear.

How to improve your AI search visibility after the audit

Fixes fall into three families, in this order of effectiveness.

• Fix what is wrong. Incorrect information about your company, a consistent name, address and business description everywhere on the web, an up-to-date Google Business Profile. It is quick, and the effect is the most direct.
• Get into the cited sources. The directories and comparison sites identified in step 4. Many accept a free or low-cost listing.
• Create the missing content. One page per important question left unanswered, written to be cited: an answer in the first two sentences, dated figures, a clear structure.

What comes next depends on the engine. For ChatGPT, I go through the levers in my article on ChatGPT SEO. For the general method, see my GEO SEO guide, and for writing passages that are easy to cite, the article on AEO. The audit tells you where you stand, GEO tells you what to do next.

The special case of local businesses

For a shop or a tradesperson, AI visibility quickly overlaps with local SEO. When someone asks an AI for a professional in a city, it relies heavily on the same signals as Google's local pack: the Google Business Profile, reviews, and consistent contact details across directories.

In practice, for a small local business, an AI visibility audit overlaps about eighty percent with a well-done local SEO audit. That is good news: the workstreams are known, proven, and they serve both channels at once.

Method mistakes that skew everything

I have seen a lot of AI visibility audits over the past year, including expensive ones. Here are the flaws that keep coming back, and that make the conclusions useless.

• Measuring from a logged-in account. This is mistake number one. The engine has remembered your previous conversations and cites you because it knows you, not because you deserve to be cited.
• A single measurement. Answers vary from one day to the next. An audit without repetition documents a moment, not a position.
• A panel written by the company alone. A team naturally phrases questions with its own industry jargon, which is not the customers' vocabulary. Always have the panel reviewed by an outsider.
• Ignoring accuracy. Many audits count citations without checking what is said. A citation with a mistake about your prices costs you more than an absence.
• Confusing traffic and citation. Being cited doesn't necessarily bring a visit. They are two distinct metrics, to be tracked separately.

If you would rather delegate, my AI visibility audit applies exactly this protocol: a panel built with you, repeated measurements on the main engines, share of voice against your competitors and a prioritized action plan.

Written by

Anthony Courtin

SEO, GEO & AI automation consultant

I help SMBs and independent professionals with their visibility (Google, ChatGPT, Perplexity) and with real-world use of AI at work. I work with companies on site to train their teams and rethink their processes. Everything I write here, I use with my clients.

Follow me on LinkedIn My YouTube channel (in French)