For SEO agencies

How to show a client whether ChatGPT, Claude and Gemini name their brand: a 7-step method for agencies (2026)

Your client asks: "Does ChatGPT recommend us?" A screenshot is not an answer. Here is the method we use to turn that question into numbers a client can read, and repeat next month.

Drafted by our AI research agents. Reviewed and approved by Jordan, founder. Published October 28, 2026. 9 min read.

Why is a screenshot not enough?

Because AI answers change. The same question can name a brand in one run and skip it in the next, and each assistant reads different pages. One screenshot shows one moment. A client needs a count over many questions, many runs and several assistants, measured the same way every month.

Even ChatGPT says so. On October 28, 2026, we asked it "how can my SEO agency show clients if ChatGPT mentions their brand". It answered mostly with a method: build a set of real customer questions, run them on a schedule, and report clear numbers. It named one tool, Surfer's AI Tracker, only as an option for larger programs. This guide is that method, made concrete.

Google describes why the answers move. Its AI features use "query fan-out": the model runs several related searches of its own before it answers. So a small change in wording can pull in different pages, and a different brand.

What are the 7 steps?

Write the buyer questions, pick 3 competitors, ask each question several times on each assistant, count the answers with fixed rules, read the cited pages, show the client the gaps, then repeat the same questions every month.

1. Write the questions your client's buyers ask. Use the category, the problem and the budget, not the brand: "best CRM for a 10 person sales team", "cheap alternatives to Salesforce". Keep 1 or 2 head-to-head questions ("HubSpot vs Salesforce for a startup"), but know that they name the brand for the assistant. A full check uses 30 questions.
2. Pick 3 competitors. The ones the client loses deals to. A score means little without someone to compare it to.
3. Ask each question on each assistant, with web search on. ChatGPT, Claude and Gemini at least. Start a fresh chat each time, so past conversations do not color the answer.
4. Ask each question more than once. We ask each one 3 times in a full report. A brand named 3 times out of 3 is solid. A brand named once out of 3 is luck.
5. Count with fixed rules. Is the brand named, as a whole word? In which place? Who is named first? Which pages are cited, and are any of them the client's own pages? Write down the exact sentences that name the brand. Do not grade the tone: read it.
6. Read the pages that win. For every question the client loses, note the page the assistant cited. That page is the work: it is where the competitor gets its credit.
7. Repeat the same questions every month. Same questions, same rules, same assistants. That is the only way a change shows up as a change, and not as noise.

Want this done for one of your clients? Get a free sample report within 48 hours, under your logo.

What does it look like on a real brand?

On October 28, 2026, we ran 10 questions on HubSpot, once each on ChatGPT, Claude and Gemini: 30 answers, all 30 returned. HubSpot was named in all 30. That looks perfect, yet it was not named first on 3 questions, and the 3 assistants read very different pages.

• Named everywhere, first on 7 of 10. HubSpot came first on 7 questions on all 3 assistants. On "best CRM for a 10 person sales team", Pipedrive came first on all 3.
• The assistants did not always agree. On "best CRM with a free plan and automation", ChatGPT and Claude put Zoho first, Gemini put HubSpot first. On "cheap alternatives to Salesforce", a question that names Salesforce, HubSpot was 4th of the 4 brands we tracked on ChatGPT, 3rd on Claude and 2nd on Gemini.
• The brand's own site was cited very unevenly. ChatGPT cited hubspot.com or its subdomains in 7 of its 10 answers, Gemini in 2, Claude in none.
• Gemini kept returning to one outside page. A US Chamber of Commerce guide to low-cost CRM tools was cited in 5 of Gemini's 10 answers (the uschamber.com site in 6), the most cited page of the run, and by neither ChatGPT nor Claude. For a client who is missing from that kind of page, that is the first lead, on that assistant.
• 82 sentences named the brand, copied word for word: 33 on ChatGPT, 29 on Claude, 20 on Gemini. The client can read how each assistant describes it, without anyone giving the tone a score.

Read it as a snapshot. One answer per question, the small default models our tool ran that day, one brand. HubSpot is a famous brand, so this is the easy case. Most of your clients will be named far less often. That is when the method earns its keep: the lost questions and the pages that win them become the plan.

What should you put in front of the client?

Five things, on a few pages: a visibility score, the brand against 3 competitors, who is named first on each question, the pages the assistants cite, and the exact sentences that name the brand. Then the agency adds its advice. The measure and the fix stay separate.

• A visibility score: the share of answers that name the brand, per assistant.
• The brand against 3 competitors: how often each is named, and in which place.
• Question by question: who is named first, on each assistant.
• The cited pages: the full address of each page, with the client's own pages flagged.
• The exact sentences: what each assistant actually says about the brand.

Say the limits out loud, too. These checks go through the assistants' official APIs with web search on, so an answer can differ from what a person sees in the app. And no one can promise a place in an AI answer.

What about Google's AI answers?

For AI Overviews and AI Mode, use Google's own tool. Search Console has a Generative AI performance report. Google says a missing report "usually happens when your site hasn't received enough impressions" in those features.

Google also warns that "No third-party tool has access to our internal ranking or AI systems." So measure Google from Search Console, and measure ChatGPT, Claude and Gemini by asking them, as above. For Google, optimizing for its AI answers "is optimizing for the search experience, and thus still SEO". The method above does not replace SEO. It shows the client where the brand stands in answers that no dashboard covers.

Should you do it by hand or with a report?

By hand works for a first look. A full Sillage Data report is 30 questions, asked 3 times on 4 assistants (ChatGPT, Claude, Gemini and Perplexity): 360 answers to read and count. Once a month, for several clients, that is when a ready-made report pays for itself.

Sillage Data makes these white-label reports for agencies. A free sample on one of your clients comes within 48 hours. A full report costs $349, delivered within 3 business days or refunded. Your logo is on it, with no Sillage Data branding. To compare other options, see the best white-label AI visibility reports for agencies.

The bottom line

A screenshot is one moment. A count over many questions, several assistants and several runs, repeated every month, is something a client can act on. Show the counts, show the pages that win, and let the agency do the fixing.

Questions agencies ask

Is a screenshot enough to show AI visibility?

No. It shows one answer on one day. Count many questions, on several assistants, several times each, with the same rules every month.

How many questions do you need to measure AI visibility?

Enough to see a pattern, not one lucky answer. We use 10 questions for a free sample and 30 for a full report, each asked 3 times on each assistant. One answer per question is a hint, not a measure.

Should the questions include the brand name?

Mostly no. Buyers ask about a category or a problem, like "best CRM for a 10 person sales team". Keep 1 or 2 head-to-head questions, but know that a question naming the brand will almost always get the brand named back.

Why do ChatGPT, Claude and Gemini give different answers?

They search the web differently and read different pages. In our HubSpot run, the 3 assistants agreed on who came first on 9 of 10 questions, yet ChatGPT cited the brand's own site in 7 of 10 answers and Claude in none.

Can an agency guarantee that ChatGPT will recommend a client?

No. Answers change from one run to the next, and no outside tool controls them. A report shows where a brand stands on a given date and which pages win the questions it loses. That is what the agency works on.

How do I measure visibility in Google's AI Overviews?

Use the Generative AI performance report in Google Search Console. It covers AI Overviews and AI Mode. Google says it may not show until a site has enough impressions in those features.

Does Google's guide cover ChatGPT and Claude?

No. Google's guide on generative AI features covers Google Search, such as AI Overviews and AI Mode. ChatGPT, Claude and Gemini have to be checked by asking them.

How often should an agency re-check?

Every month, with the same questions, the same assistants and the same rules. That way a change in the numbers is a real change, not a different test.

How we tested and sourced this article

The HubSpot example is a real run of our tool, Brand AI Answers, on October 28, 2026: 10 questions, one answer each on ChatGPT (gpt-6-luna), Claude (claude-haiku-5-5) and Gemini (gemini-2.5-flash-lite), through their official APIs with web search on. Apify run mvmeTWehw4xZ9SK0i. All 30 answers came back. HubSpot is a public brand used as an example, not a client.

Google: Optimizing your website for generative AI features on Google Search and the Generative AI performance report. The ChatGPT answer quoted at the top comes from our 16-answer test of October 28, 2026, described in our comparison of white-label reports. We are not affiliated with HubSpot, OpenAI, Anthropic or Google.

Related reading

• ChatGPT vs Claude vs Gemini: do they recommend the same brands?
• How to track your brand in ChatGPT answers