How do you check whether ChatGPT, Gemini and Perplexity mention your brand?
Short answer
Write 15–20 questions your customers really ask without naming your brand, put them to several AI assistants in a clean session, record every brand named and every source cited, and repeat the same set weekly. One answer means little; the share of answers that mention you over several runs is the real signal.
Why can't you just ask ChatGPT about your company?
Asking "What do you know about [your brand]?" tells you whether the model recognizes your name, which is useful, but it doesn't tell you whether customers will meet you. Customers don't ask about you; they ask about their problem: "where can I buy a laptop with next-day delivery", "which accounting firm in Kraków works with IT companies". The real test is whether you appear in answers to those questions.
Step 1: write the questions your customers ask
Build a set of 15–20 questions that cover:
- Category questions: "best online electronics store in Poland".
- Problem questions: "laptop for video editing under 4000 zł where to buy".
- Comparison questions: "MediaShop or TechNova, which is better for returns".
- Local questions: "web design studio in Chisinau for e-commerce".
- Price questions: "cheapest place to order a 65-inch TV with installation".
Rules: don't include your brand name, write in your customers' language, and keep the wording natural. Sales and support teams are the best source of real phrasing.
Step 2: choose the assistants to test
Different assistants give different answers, so test more than one:
- ChatGPT, with search on and off.
- Gemini, which uses Google Search for many questions.
- Perplexity, which always searches and shows sources.
- Claude and Copilot, if your audience uses them.
For automated monitoring, models available through APIs are used as indicators: for example Gemini with Google Search grounding, plus open models such as Llama or Mistral to represent answers from model memory.
Step 3: remove personalization
Assistants remember your previous chats and may know where you work. Test in a logged-out or temporary chat, without custom instructions, and set the location to your market when the tool allows it. Otherwise you may see your brand only because the assistant knows you are interested in it.
Step 4: record more than "yes or no"
For each answer write down:
| Field | Why it matters |
|---|---|
| Brands named, in order | Position matters: the first name gets most attention |
| Whether you are named | The base of your share of voice |
| How you are described | Wrong prices or outdated facts need fixing |
| Sources cited | Shows which sites shape the answer |
| Date and model | Answers change over time and between models |
A spreadsheet works for a first look. For weekly tracking across several models, automation saves hours.
Step 5: repeat and look at the trend
AI answers are not deterministic. The same question can produce a different list tomorrow. That's why a single test proves nothing: run the full set at least weekly and compare percentages, not individual answers. If you are named in 4 of 20 answers this week and 9 of 20 three weeks later, that change is meaningful. The metric is explained in AI share of voice.
Step 6: read the results
Typical patterns and what they mean:
- You are never named, and competitors are. Look at the cited sources: competitors are probably present on review and ranking sites where you are not. See where AI gets its answers.
- You are named only by assistants without search. Your brand is known, but your current pages may be blocked or hard to read. Check robots.txt rules for AI bots.
- You are named with wrong facts. Fix the facts on your site and on third-party profiles, and make them consistent.
- Your own site is never cited as a source. AI reads about you elsewhere. Check whether your pages have readable text and clear facts (JavaScript and AI crawlers).
Can this be automated?
Yes. A monitoring setup stores your question set, queries several models on a schedule, extracts the brands from each answer, calculates shares and positions, and lists the sources. Rekomo does exactly this and also flags new competitors that start appearing in answers. You can start with the free technical audit and add weekly monitoring when you are ready.