How to know if your brand appears in ChatGPT (and Google's AI answers)

By Gecko Studio · updated October 2026

This is, literally, the question that reaches us most from AI answers. We have copied it as people type it, because it shows up in our own Search Console (originally in Spanish; translated here):

how do I know if my brand is still visible when buyers search in ChatGPT instead of Google?

And its variants, all real:

• How do I know if competitors are being mentioned in AI answers instead of us?
• How can I check whether my small business appears in AI search results such as ChatGPT?
• How can I tell whether our brand is being excluded from AI-generated recommendations?
• How to check whether AI chatbots mention my business today.

The short answer is that you can check, but you cannot measure. They are two different things, and almost everything sold right now confuses one with the other. Below are the four checks that work, in order of what they cost, and at the end what you will not be able to know however much you pay.

Before we start: what this is not

We are not going to talk about "GEO" or "AEO" as if they were new disciplines. Google has said explicitly that they are not, and we covered it in Google says GEO and AEO are myths. Appearing in an AI answer has no lever of its own: it depends on being readable, on there being something to cite, and on other people talking about you. What is new is how hard it is to know whether it is happening, and that is what this page is about.

1. Search Console is already showing you (free, and today)

This is the check almost nobody does, and it costs nothing: your usual performance report contains queries that nobody typed into a search engine. Whole conversational sentences, prompts pasted as they were, and loose replies such as "give me the direct links".

If your site appears for queries like these, you are entering AI conversations.

Search Console → Performance → Queries → Add filter → Query → Custom (regex), and paste this to see queries of twelve words or more:

^([^ ]+ +){11,}[^ ]+$

How to read the result, what to strip out first and why twelve words rather than sixteen are in the full method with the three filters.

A caution on interpretation. Ten conversational queries does not mean you had ten. Google suppresses low-volume queries, and a conversational query is rare by definition, so many never appear. We measured it on one site in our portfolio by comparing the real total against the sum of the per-query table: between 53% and 56% is lost. What you see is a floor. It is good for confirming that you appear and for reading how people ask. It is not good for counting.

2. Ask the models, but with a method

It seems obvious and it is where most people go wrong: open ChatGPT, search for your own brand name, see that it appears, and call it done.

That checks nothing. Four rules to make the test count:

• Logged out. Your own history shapes the answer. Use an incognito window or a clean account.
• Do not search for your brand: search for your problem. Asking "what is [your brand] like?" only proves that the model knows who you are. What you want to know is whether it recommends you when someone describes the need without naming you: "I need an agency to handle the SEO of an online shop in Spain".
• Several wordings, over several days. The same question gets different answers. One test is an anecdote; ten wordings repeated over two weeks start to be a trend.
• On the ones that matter to you. ChatGPT, Google (AI Overviews and AI Mode), Perplexity, Claude and Copilot do not answer in the same way or draw on the same sources.

For each answer, note whether you appear, at what position in the list, and who it cites as a source. That last point is often the most useful of all: if in your sector the AI always cites the same three directories or publications, you know where you are missing.

3. Check that they can read you

You can do everything right and still be invisible because of a switch your CDN turned on without telling you. It is common: Cloudflare enabled blocking of AI crawlers by default on hundreds of thousands of sites, and that block sits above your robots.txt, so your file says one thing and the server does another.

The three concrete steps (request your real robots.txt, check the CDN dashboard, and look at your logs by user-agent) are in If ChatGPT does not mention you, you do not exist, and the detail of what they crawl and what they ignore is in Your website and AI bots.

Do this before investing in content. If they cannot read you, nothing else matters.

4. Monitoring tools: what exactly you are buying

There are paid products that run your questions against several models on a schedule and build a dashboard showing how things change. We use one for client work.

It is worth knowing how they work inside, because it changes what you can expect from them: they have no privileged access to ChatGPT. They do exactly what step 2 describes, automated and at scale. We know this for certain because their questions reach us: in our own Search Console there are queries with the template left unfilled, exactly like this:

how does {{brand}} differ from its competitors in the Spanish market?
<context> company url: https://example.com  market: spain </context>

That is a monitoring tool doing its rounds. On our own site, which is an industry site, almost three in ten conversational queries are machines of this kind. On the client sites we run, it is one in eight.

The practical conclusion: a tool like this saves you the manual work and gives you a historical series, which is no small thing. What it does not give you is the truth about how many people saw you. It gives you how a sample changes over time.

What you will not be able to know (and anyone who says otherwise is lying)

• How many times ChatGPT mentioned you this month. That figure does not exist. OpenAI does not publish it and no tool has it.
• How much traffic AI Mode sent you, as a reliable number. Search Console gives impressions in the generative AI report, but without the queries; and the query table is anonymised for more than half. You cannot build a business figure on that.
• Why the model chose your competitor. You can guess from the sources it cites. There is no auditable explanation.

If someone sells you an "AI visibility" report with percentages to the decimal and month-by-month trends, what they are showing you is how their own questions change. That can be useful. It is not a market measurement.

So what do I do with the result?

The same as we do with a client, in order:

1. Check that they can read you. Step 3. It is the only thing that can be completely broken.
2. Pull from Search Console how people really ask. Step 1. That language, word for word and without "professionalising" it, goes into your H2s, your FAQs and the opening lines of each page. People talk to AI the way they talk, and your site is almost never written that way.
3. Look at whom the AI cites in your sector. Step 2. That is where the work outside your site lies: being where the models go to look.
4. Repeat in three months. Without a series there is nothing to read. A single snapshot says nothing.

And something that is easy to forget: what gets you cited is still, largely, the usual. Content of yours with concrete data, other people talking about you, and being understood. We develop this in Trust as a ranking factor.

If you would rather we checked for you, it is part of what we look at in an SEO audit. We tell you whether they can read you, what people are asking about you today without your knowing, and who appears in your place.