No tool can tell you exactly how often ChatGPT, Gemini, or Perplexity cite a brand: answers change from one request to the next. You can still measure honestly with a fixed set of prompts repeated over time, alongside the official data in Google Search Console, Bing Webmaster Tools, and GA4. What you get is a trend, not an absolute score.
Why a single test tells you nothing
Google writes that AI Overviews and AI Mode may use different models and techniques, so the responses and links they show will vary, and that AI Overviews only appear when its systems judge them additive, so they often don’t trigger. For the other assistants we found no official data on variability, but the same caution applies: the same question asked twice can return different answers.
It is reasonable to expect differences by language, location, account history, and date too. That is a methodological precaution, not a documented fact, which is why the method below builds in repetition and controlled conditions.
How to build a fixed prompt set
You need 15 to 30 questions, written once and never edited, split into categories:
- Category. “What are the best branding agencies in Milan?” In other words, questions about what you do, with no names.
- Problem. The questions a customer asks before knowing you exist.
- Comparison and choice. “How to choose a supplier for…”, “what to ask when hiring…”.
- Brand. A few questions with your name, to check what is said about you.
Category and problem questions are the most useful. If the brand name is already in the prompt, the answer will almost certainly contain it and the test measures nothing.
How to run the tests
- Use a private window or a test account with no history tied to your company.
- Repeat each question on different days, for example three times over two weeks, and repeat the whole cycle every month.
- Test every assistant you care about, in the language your customers use.
- If web search is switched on, note it: an answer with sources and one without are not comparable.
In a spreadsheet, log for each test: date, assistant and model, language, prompt, brand presence (yes/no), relative position among the names cited, sources cited, link to your page (yes/no), and notes. Save a screenshot too. At the end, count how often you appear per category, not any single result.
Mention or citation: not the same thing
A mention is your name in the text of the answer. A citation is your page listed as a source, usually with a link. Log the two separately. Perplexity, for example, says that each answer includes numbered citations linking to the original sources: there, citation is part of the format. A mention without a link tells you the name circulates; a citation tells you which of your pages was used, and that is the one you can improve.
What official tools can measure
Google Search Console. According to the documentation, sites appearing in AI features are included in overall search traffic and reported in the Performance report, under the Web search type. The counting rules are in the Search Console help: a click on an external link in AI Overviews or AI Mode counts as a click, and all links in an AI Overview share the same position. The page we reviewed doesn’t describe a way to separate this data from other Web traffic. Google’s guide to optimizing for AI search also points to a dedicated generative AI performance report: check whether it is available for your property.
Bing Webmaster Tools. There is an AI Performance report, introduced as a public preview: it shows total citations, average cited pages, grounding queries (the phrases the AI used to retrieve content), and citations per page, across Microsoft Copilot, AI summaries in Bing, and select partner integrations.
GA4. OpenAI states that ChatGPT adds utm_source=chatgpt.com to referral URLs from search. In GA4 you can therefore look at visits from that source. For other platforms we found no documented convention: check your acquisition report to see which sources actually appear. Only clicks are counted: someone who reads the answer and doesn’t click leaves no trace.
As for third-party tools, Google warns to be wary of those that promise ranking success or claim to use internal Google metrics.
Why outside sources matter
Answers that use search draw on web pages, not only the brand’s own site. Perplexity says it gathers information from authoritative sources like articles, websites, and journals. Google explains that AI Overviews and AI Mode may use query fan-out, running multiple related searches across subtopics and sources. When you log the cited sources, look at which third-party sites show up for your categories too: they are the context in which the brand gets described. Stay cautious: Google says chasing inauthentic mentions around the web isn’t as helpful as it might seem.
Mistakes to avoid
- Measuring once and drawing conclusions.
- Using prompts that already contain the brand name.
- Changing the questions every cycle: you lose the comparison over time.
- Confusing mention and citation, or counting only the favorable cases.
- Reading a result as a guarantee: no engine promises a citation. OpenAI itself advises not blocking OAI-SearchBot to be cited and linked, without promising inclusion.
Before measuring, make sure your pages are reachable by the right crawlers: we cover that in how to get cited by AI assistants. For the big picture, see the differences between SEO, GEO, and AEO.
Magnet handles SEO, GEO, and AEO even for sites built by others: if you want to set up monitoring or understand what assistants read on your site today, it’s all on the SEO, GEO, and AEO page.