Does your company appear in ChatGPT's answers? How to measure it

Yes, you can find out whether ChatGPT recommends your company, for free, in thirty minutes a month. Djtal has been applying this protocol to its own visibility since June 2026: ten fixed questions, rerun at regular intervals, with dated results. This article gives the full method and our real figures, including the unflattering ones.

The invisible meeting

In June, a business leader in French-speaking Switzerland got in touch with us. His search had begun away from Google and from word of mouth: he had asked an AI to list qualified firms in artificial intelligence. We never saw his question or the answer. If an engine had left us out that day, the lost opportunity would have stayed invisible to us.

This is the blind spot of 2026. A growing share of B2B buying decisions starts in a conversation with ChatGPT, Perplexity, Copilot or Gemini, and those answers leave no trace in your analytics unless someone clicks. The only way to see them is to run them yourself.

The method: ten questions, six steps, thirty minutes

No paid tool is needed. The protocol comes down to six steps:

1. Fix a set of 10 to 20 questions that a customer would ask, in three families: your exact speciality (“who can connect an AI agent to a Zoho CRM in Switzerland?”), the generic side of your market (“which AI agency suits an SME in French-speaking Switzerland?”) and your brand (“how good is company X?”). The set never changes, and that is what makes the runs comparable.
2. Neutralise personalisation with a private window, a logged-out session or a temporary chat, memory switched off and web search switched on. Otherwise the measurement reflects your own history and your market stays out of the picture.
3. Standardise the conditions: a Swiss IP address (answers vary by country), the same engines, the same time of the month.
4. Record three things per question: cited, mentioned or absent; who comes first; which sources the engine displays. Those sources are the most valuable data, because they decide the recommendations.
5. Rerun the set every month and follow the trend. Answers vary from one run to the next, so one snapshot proves little; three runs start to show which way things are moving.
6. Cross-check with the free signals: the AI Performance report in Bing Webmaster Tools (citations in Copilot), Search Console, referral traffic from AI assistants in your analytics tool, and your server logs, where AI crawlers identify themselves by name.

Our figures, warts and all

We ran two passes: 22 June on Perplexity and 6 July on ChatGPT. Djtal is cited in two out of ten questions. We are absent from the answers on “AI agency in French-speaking Switzerland”, “AI consultant” and “AI audit”, even though those pages exist on our site and we already deliver these services to clients.

The detail tells you more than the score. On “best Zoho partner in French-speaking Switzerland”, we come out first, with our local listing and phone number shown. On “who can integrate an AI agent into my Zoho CRM?”, ChatGPT placed us first on 6 July (“recommended as a priority”) and cited our client references. Two weeks earlier, Perplexity had ranked us second. Between the two passes, the engines absorbed what we had published and began serving it back to their users.

What the measurement teaches: the gradient rule

Our two passes tell the same story: the likelihood of being recommended by an AI is proportional to the precision of the question. On a narrow question that falls within your speciality, the engine readily recommends the specialist, since it has a verifiable reason to name you. On a generic question, it compiles directories, reviews and press articles, and credibility there belongs to third parties. Your own site carries almost no weight.

For a business leader, two consequences follow directly:

• Your speciality is your best chance of being recommended. Name it precisely, everywhere, in the terms a customer would use. A specialist position can be won in a matter of weeks. A generalist position may stay out of reach for good.
• Generic questions are settled outside your website. Verified customer reviews, press mentions and consistent directory listings make up the third-party layer that engines consult before vouching for anyone. Even a perfect site, without third-party proof, remains a candidate that engines pass over.

Frequently asked questions

Are there free tools to measure your visibility in ChatGPT and other AI engines?

Yes. Bing Webmaster Tools publishes an AI Performance report that shows how often Copilot cites your pages. Google Search Console’s generative AI performance reports show how often your URLs appeared in generative AI features such as AI Overviews and AI Mode. Your analytics tool can isolate traffic coming from chatgpt.com or perplexity.ai. None of these signals requires a subscription, and the manual pass described here complements them.

How often should you measure?

One pass a month is enough, provided you keep the same set of questions. Engines’ answers naturally vary from one day to the next, so what counts is the trend over two or three months.

Doesn’t traditional SEO already cover all this?

The two disciplines complement each other. Generative engines pick out passages and pieces of evidence, and how well a page ranks counts for less in that choice. They rely heavily on third-party sources (reviews, press, directories) that traditional SEO barely addresses. A good Google ranking helps, but the recommendation is decided elsewhere.

---

You can run your first pass tonight, with no budget: ten questions a customer would ask, a private window and a scoring grid. In thirty minutes you will know which names your market hears recommended instead of yours. It is the right place to start if you want to appear in those answers.