You don't need a dashboard to find out how visible your brand is in AI search. You need ten minutes and access to ChatGPT, Gemini or Perplexity. The prompts below aren't generic "what do you know about us" questions — each is built so that a weak, missing or wrong answer is the likely outcome for most companies, because that's what actually shows you where the gaps are. Run them, copy the exact wording, and swap in your own company name, category and city.
1. "What is [Company]?"
The baseline entity-recognition check. A strong answer names your actual category, what you sell and who you serve, in one or two correct sentences. A weak answer is vague, outdated, or confuses you with a differently-named company — a sign the AI model has no confident, consistent picture of your brand as an entity at all.
2. "Who are the best [your category] companies in [your city or country]?"
The category-leadership check. Most companies don't appear here, even ones that would appear on page one of Google for the same query — appearing in an AI's unprompted "best of" list requires a level of third-party validation that ranking well in search doesn't.
3. "I'm choosing between [Your Company] and [a named competitor] — which do you recommend, and why?"
The direct comparison. This is the prompt most likely to go badly: AI models frequently recommend the competitor, sometimes citing reasons that are outdated or simply wrong about your own company. If it happens, note the specific reason given — that's the exact gap to close.
4. "What do people say about [Company] online?"
The reputation and third-party-signal check. If the answer is thin or generic, it usually means the AI has little independent review or discussion content to draw on for your brand — the same category of source (reviews and user-generated content) that accounts for the largest single share of what ChatGPT actually cites.
5. "What are the downsides or limitations of [Company]?"
The honesty-pressure test. AI models are more willing to name negatives than most companies expect, and sometimes name ones that are outdated, exaggerated, or based on a single old complaint treated as representative. If a negative appears that isn't accurate anymore, that's specific, current-content work to do.
6. "Can you recommend a [your category] provider for [a specific use case]?" — without naming your company
The unprompted-recall test. Don't mention your own brand in the prompt. This is the closest simulation of how an actual prospect would ask, and it's the prompt most companies fail hardest, because it removes the advantage of the AI simply repeating your name back to you.
7. "What sources did you use to answer that?"
Ask this as a follow-up after any of the prompts above. An AI that can name a specific page, review site or publication is telling you exactly which source earned the citation. An AI that can't name a source — or invents one — is telling you the previous answer had no real grounding, which matters more than whether that answer happened to sound positive.
8. "Is [Company] a legitimate and trustworthy business?"
The trust-verification check, particularly relevant for regulated, financial or high-consideration categories. A confident "yes" with specifics (years in business, credentials, verifiable claims) signals real Entity Authority. A hedge — "I don't have enough information to confirm that" — is a direct, quotable signal of a trust gap.
9. "What does [Company]'s [specific product or service] cost?"
The commercial-accuracy check. Pricing is exactly the kind of fact that goes stale fastest in AI training data and cached web content — expect outdated numbers even from companies that keep their own pricing page current, and treat any answer here as a signal about content freshness, not just pricing.
10. "If someone searches for [your category] + [your strongest benefit], which companies come up, and why?"
The broader visibility check across the language your actual customers use, not just your own brand name. Run it with two or three different benefit phrasings — "fastest," "most affordable," "best for small business" — since AI models frequently surface a different shortlist for each variant of the same underlying need.
Reading your results
Score each prompt roughly as a pass, partial or fail. A handful of fails is normal, not a crisis — it's a prioritized list. Missing or wrong entity facts (prompts 1, 8, 9) point to Entity Clarity work: structured data and consistency fixes. Weak recall and comparisons (prompts 2, 3, 6, 10) point to Entity Relevance and content depth. Thin reputation signal and no citable sources (prompts 4, 7) point to Entity Authority — third-party mentions, reviews and citations you don't fully control but can systematically build.
This is a snapshot, not a system. Run it again next month and the answers may already have moved — for better or worse — as models update and as your content and third-party footprint change. If you'd rather have this running continuously instead of manually, a free AI visibility check runs the same kind of test against your brand and shows you where the gaps are today. For the deeper framework behind why these gaps exist, see how Brand Entity Equity is built, or start with the full step-by-step guide to checking what AI says about you.