Most businesses answer “Does our brand show up in ChatGPT?” by guessing: one or two searches are run, the answer is read, and the result is summarized in one sentence. But AI answers may not be the same every time; a result from a single attempt is a snapshot, not a measurement. Below, I explain step by step the measurement method I use for brands.
Why is measuring hard?
In classic search there is a single answer such as “you rank third.” In AI systems, the same question can get different answers at different times, and each platform has its own sources and behavior. That is why measurement has to be repeatable: same questions, same platforms, same recording format.
1. Building the question set
The foundation of measurement is a question set prepared specifically for your brand. It consists of questions your customer could genuinely ask before buying:
- Discovery questions: “Which brands do you recommend for [category] in Istanbul?”
- Comparison questions: “[Brand A] or [Brand B]?”
- Problem questions: “What should I do about [problem]?”
- Brand questions: “What is [brand name], is it reliable?”
In an audit I use more than 40 queries; coverage matters more than the number: which question would lead someone who has never heard of the brand to you?
2. Same platforms, same conditions
The questions are run on ChatGPT, Gemini, Claude, Perplexity and Google AI Overview. The same platforms are used in every measurement so that month-to-month comparison is possible. The platforms' own updates can change results; this shows why measurement should be done regularly, not once.
3. Evidence: screenshots
Every answer is recorded with a screenshot. That way the report presents evidence, not opinion: saying “your brand was mentioned in this question with this sentence” becomes something that can be discussed together with the screenshot.
4. Five measures
| Measure | Question |
|---|---|
| Mention rate | In how many queries does the brand name appear? |
| Mention context | As a recommendation, in a list of options, or in a negative comparison? |
| Source citation | Is your site or your competitor's site cited as the source in the answer? |
| Competitor position | Which brands are recommended in your place? |
| Incorrect information | Does the model give outdated or incorrect information about your brand? |
5. Repeat every month
The first measurement is the starting point. The real value comes from repeating the same set every month: in which question did you appear, in which did you lose, where did a competitor rise? After three measurements you start to see a trend. That is why I offer the monthly tracking service for a minimum of three months.
Google side: Search Console
The question set method also applies to platforms other than Google; but for how much your site appears in Google's own AI surfaces (AI Overviews, AI Mode), Google tells you directly. According to Google's help pages, performance from AI features is included in the “Web” search type in the Search Console Performance report; from 2026 a dedicated performance report for generative AI features is also offered (opened to all sites on 31 August 2026). This report mainly provides a breakdown by impressions, page, country, device and date. That is why Search Console data is not a replacement for question set measurement but its complement.
What a measurement does not tell you
- On its own it does not say how much of your sales come from AI; for that, the order source must be tracked separately.
- It does not guarantee future results; models are updated.
- The result from a single platform says nothing about the others.
Frequently asked questions
Can I measure it myself by asking a few questions?
To get an idea, yes. To make decisions, not enough: single attempts can be inconsistent. You need multiple questions, multiple platforms and repetition.
Sources
- Google Search Central Blog: Search Console generative AI performance reports (2026)
- Google: AI features and your website
Sources checked on 1 October 2026; official pages may change over time.
