Direct answer
What prompts should I use to check my brand's AI search visibility?
Run a small, repeatable set of prompts — category questions, comparison questions, and direct brand questions — across ChatGPT, Perplexity and Google's AI mode, then record whether your brand appears, how it's described, and which sources get cited. This is a manual sampling method, not a scientific measurement of total AI visibility; treat one session's results as a snapshot, not a verdict.
01
Test category and comparison prompts, not just your own brand name — most buyers don't type your name first.
02
Record which URLs the assistant cites, not just whether your name appears in the text.
03
Re-run the same prompts periodically; a single session is a sample, not a trend.
Before you subscribe to an AI-visibility monitoring tool, you can run this test yourself, for free, in about 30 minutes.
Replace the bracketed placeholders with your actual category, brand name and competitors, run each prompt in ChatGPT, Perplexity and Google's AI mode, and record what comes back. Don't just test your own brand name — most buyers don't type it first.
Before you start
Assistants can give different answers for the same prompt on different days, in different chats, or in different regions. One test session is a sample, not a measurement of your total AI visibility — run these prompts more than once, ideally over a few weeks, before drawing conclusions.
1–4. Category & discovery prompts
These mimic a buyer who doesn't know your brand name yet — the hardest and most valuable place to show up.
"What are the best [your category] tools for [specific use case or industry]?"
Look for: Whether your brand appears at all, where in the list, and whether the description of what you do is accurate.
"I'm a [buyer persona] looking for software to solve [the problem you solve]. What should I consider?"
Look for: Whether the assistant frames the problem the way you'd frame it, and whether your brand — or a competitor — gets named as an example solution.
"What's the difference between [your category] and [an adjacent category]?"
Look for: Whether the assistant's category definition matches how you position yourself, since a mismatch here can mean you're being filed under the wrong mental category entirely.
"What are common mistakes companies make when choosing a [your category] tool?"
Look for: Whether any named mistake maps to a criticism you've heard from real prospects — this often surfaces objections worth addressing on your site.
5–8. Comparison prompts
These mimic a buyer who already has a shortlist and is deciding between named options.
"Compare [your brand] vs [competitor A] vs [competitor B]."
Look for: Whether the comparison is factually accurate, which sources it cites, and whether any claimed differentiator is actually true today.
"Which is better for [specific use case], [your brand] or [competitor]?"
Look for: Whether the assistant can even answer — a vague or hedged response often means neither brand has clear enough public information about that specific use case.
"What are the pros and cons of [your brand]?"
Look for: Whether listed cons are outdated, inaccurate, or based on old reviews — this is a common source of stale-information corrections worth pursuing.
"Is [your brand] good for [a specific company size or industry]?"
Look for: Whether the assistant has enough specific information to answer confidently, or defaults to generic hedging because your site doesn't state this explicitly.
9–12. Direct brand prompts
These test the baseline: what does the assistant actually know when asked about you by name?
"What does [your brand] do?"
Look for: Whether the description matches your current positioning — brands that have repositioned recently often get described using stale, outdated language.
"How much does [your brand] cost?"
Look for: Whether pricing is current, and whether it cites your actual pricing page or a third-party listing that may be wrong.
"What do people say about [your brand]?"
Look for: Which review sites or forums get cited, and whether the summarized sentiment matches what you see in your own review monitoring.
"Who are [your brand]'s competitors?"
Look for: Whether the competitor set matches how you actually see the market — being grouped with the wrong competitors is itself a positioning signal worth acting on.
How to log results so they're actually useful later
A spreadsheet with these columns is enough to turn a one-off test into a repeatable check:
Date and assistant
When you ran the prompt and which system (ChatGPT, Perplexity, Google AI mode, Gemini) answered it.
Mentioned: yes / no
Whether your brand appeared anywhere in the response, even without a citation link.
Cited: yes / no, and which URL
Whether a specific page of yours was linked or referenced as a source, and which page it was.
Accuracy notes
Anything stated about you that's outdated, wrong, or based on an old version of your product or pricing.
Sources cited for the whole answer
The other domains referenced in the response — these are often the pages actually shaping the assistant's view of the category.
For a more structured version of this — including how to calculate citation rate and share of voice against named competitors — see LLM Citation Auditing & Brand Share of Voice.
What this manual test can't tell you
A dozen prompts run once is not a statistically representative sample of everything ChatGPT or Perplexity might ever say about you. A missing mention in this session is an observed gap at this point in time — not proof that you're permanently excluded, and not proof of a specific technical cause.
If the pattern holds across repeated tests, the next step is diagnosing why — which usually starts with checking whether AI crawlers can access your site at all. That's covered in How to Get Your Brand Cited by ChatGPT, Perplexity & AI Overviews.
Found a gap in these prompts?
Find out whether it's an access problem or a content problem.
Run Growthract's free diagnostic to check whether AI crawlers can actually reach and parse your site before assuming the content itself is the issue.
AI visibility prompt-testing FAQs
How often should I re-run these prompts?
Monthly is a reasonable starting cadence for most B2B SaaS teams — frequent enough to catch drift, infrequent enough to stay manageable without dedicated tooling.
Why did the same prompt give a different answer yesterday?
AI assistants can vary responses across sessions, accounts, and model updates. That variability is exactly why one test session should be treated as a sample rather than a fixed measurement.
Should I use my personal ChatGPT account or an incognito session?
Testing in a logged-out or fresh session reduces the chance that your own chat history or saved preferences skew the answer — closer to what a new prospect would actually see.
Continue exploring
Related insights
Case study
An illustrative strategy showing how a Series B fintech could improve citation coverage from 10% to 34% across a controlled set of high-intent ChatGPT and Perplexity prompts.
See the case study →Insight
A practical framework for improving your chances of being discovered, understood, sourced and cited across ChatGPT, Perplexity and Google AI search without relying on fake citation hacks.
Read the insight →Insight
A #1 Google ranking does not guarantee your brand appears in AI Overviews, AI Mode, ChatGPT or Perplexity. Why the visibility paths differ, and what to measure next.
Read the insight →