On this page
Key takeaways
- Use questions your buyers ask that do not include your brand name. Those show whether AI recommends you on its own.
- Ask each question in a fresh chat, ideally logged out or in a temporary chat, so past conversations do not shape the answer.
- Check more than one AI tool. ChatGPT, Gemini, Claude and Perplexity often name different companies.
- Ask each question more than once. Answers vary, so count how often you are named, not whether you were named once.
Step 1: Write the questions your buyers ask#
Start with 10 to 20 questions. Write them the way a buyer would type them, including their role, company size or industry where it matters. Mix the stages of buying:
| Stage | Example question |
|---|---|
| Problem | How can a small marketing team track pipeline from LinkedIn ads? |
| Learning | What should I look for in a B2B attribution tool? |
| Comparing | What are the best B2B attribution tools for a 50-person SaaS company? |
| Choosing | Is [your company] a good fit for a 50-person SaaS company? |
Keep most questions free of your company name. "Best tools for X" questions show real visibility. A question that names you will almost always mention you, so it tells you how AI describes you, not whether AI recommends you.
The free Buyer-questions worksheet has patterns for each stage.
Step 2: Set up a fair test#
Small details change AI answers, so keep the test clean:
- Use a fresh chat for each question. In ChatGPT, a temporary chat avoids memory from earlier conversations.
- Be logged out where possible, or use an account with no history about your company. If you have asked ChatGPT about your own company many times, it may remember.
- Note the mode. ChatGPT with search turned on reads the web live and cites sources. Without search, it answers from what it learned in training. Both matter, and they can differ.
- Write down the date. Answers change over time as tools and models are updated.
Step 3: Ask each AI tool#
Ask the same questions in:
- ChatGPT (chatgpt.com)
- Gemini (gemini.google.com), and check Google's AI Overviews in normal search too
- Claude (claude.ai)
- Perplexity (perplexity.ai)
Each tool uses different models, sources and search providers, so being named in one does not mean being named in the others.
Step 4: Record what matters#
For each answer, record:
- Named or not. Is your company in the answer?
- Position. First, middle or last? The first names get the most attention.
- Who is named instead. List every rival mentioned. Patterns appear quickly.
- How you are described. Is the description right? Wrong facts, old pricing or the wrong category are common and fixable.
- Sources cited. Perplexity and ChatGPT search show links. These are the sites shaping the answer in your category.
Step 5: Repeat, then count#
AI answers are not fixed. The same question asked twice can return a different list. So ask each question at least twice per tool, and report a share rather than a yes or no:
Named in 9 of 24 answers (38%) across four AI tools.
With small numbers, treat the result as a range rather than an exact figure. We explain why in How to measure AI visibility.
Step 6: Turn results into actions#
| What you see | What it usually means | First fix |
|---|---|---|
| Not named on any "best tools" question | AI does not connect you to the category | Say your category plainly on your homepage and profiles; publish a use-case page |
| One rival named everywhere | That rival has strong pages and mentions | Write an honest comparison page against them |
| Named, but described wrongly | Old or mixed facts across the web | Fix your About page, schema and profiles so they match |
| Named in ChatGPT search, not without search | The live web knows you; training data does not yet | Keep earning mentions; this usually catches up over time |
| Cited sources never include you | You are missing from the sites AI reads | Get listed on the review sites and roundups that are cited |
More on this in Why AI recommends your competitor instead of you.
A faster way#
Doing this by hand takes a few hours for 20 questions across four tools, and longer to repeat each month. Our free AI visibility check asks five buyer questions about your website and shows who is named, in about 2 minutes. The Deep report runs your questions twice on all four tools with every answer shown word for word, and Live AI Monitoring repeats it every week.
Questions and answers
Does ChatGPT give everyone the same answer?
No. Answers vary between runs, between users with different chat history and memory, and between search on and off. That is why a fair test uses fresh chats and counts results across several runs.
Should I include my brand name in the questions?
Include a few, to check how AI describes you. But most questions should not name you, because those show whether AI recommends you without being prompted.
How often should I check?
Monthly is enough for a manual check. If AI search is an important channel for you, weekly tracking shows trends sooner and catches when a rival moves ahead.
Why does Perplexity name us but ChatGPT does not?
They use different models, search providers and sources. Perplexity searches the web for most answers and cites them, so a single strong page or listing can get you named there sooner.
B2B marketers who build tools to measure how ChatGPT, Gemini, Claude and Perplexity talk about companies. About us