Does ChatGPT Recommend Your Product? Here's How to Find Out in 20 Minutes
The short answer: open ChatGPT, Claude, Perplexity and Gemini in a logged-out or temporary session, ask each one the questions your buyers actually ask ("best [your category] for [your customer type]", "alternatives to [your biggest competitor]"), and record whether you're mentioned, where you appear in the list, and how accurately you're described. Run each question two or three times, because AI answers change between runs. The whole exercise takes about 20 minutes and most companies are surprised by what they find.
Here's how to do it properly, including the mistakes that give most people a false reading.
Why you can't just ask ChatGPT once and call it done
The obvious approach is to open ChatGPT, type "best project management software" and see if you show up. Most founders have done this at least once, usually after a conference talk scared them.
The problem is that this test lies to you in two directions.
First, if you're logged in, ChatGPT knows who you are. It has your chat history, possibly a memory of your company, and it personalises accordingly. If you've ever discussed your own product with it, you are far more likely to appear in its answers than you are in your prospect's answers. You're not testing your visibility, you're testing your own echo.
Second, AI answers aren't fixed. Ask the same question three times and you'll get three slightly different shortlists. A single run tells you almost nothing. You need a few runs across a few engines before a pattern emerges.
So the DIY version needs a bit more rigour than a quick chat. Not much more. Here's the process.
Step 1: Write down the questions your buyers actually ask
Don't test your brand name. Anyone can get ChatGPT to describe their own company. The test that matters is whether you appear when a buyer asks a question that doesn't mention you.
You want five to eight questions across these four types:
Category questions. "What's the best [category] for [customer type]?" Be specific about the customer. "Best CRM" is a different battle to "best CRM for a 30-person recruitment agency", and the second one is closer to how real buyers ask.
Competitor questions. "What are the best alternatives to [your biggest competitor]?" This is the highest-intent question in the set. Someone asking it has budget and a shortlist forming.
Comparison questions. "[Competitor A] vs [Competitor B], which is better for [use case]?" You're checking whether AI ever introduces you as a third option, which well-cited brands get regularly.
Problem questions. "How do I [solve the problem your product solves]?" Earlier stage, but this is where category leaders get named before buyers even know the category exists.
Write these down before you start. You'll reuse the exact same wording every time you run the audit, which matters for comparing results month to month.
Step 2: Run them where your buyers run them
Your prospects aren't all using ChatGPT. Run your questions across the four engines that show up most in B2B buying research:
ChatGPT, in a temporary chat (click the model name, select "Temporary chat") so memory and history don't skew the answers
Claude, in a fresh conversation with no project or custom instructions attached
Perplexity, logged out if possible, because it shows its sources, which you'll want in Step 3
Gemini, in a fresh chat
Ask each question in each engine two or three times. Yes, that's up to 90-odd runs if you do the full set. In practice, most people test their three most important questions across all four engines and get a reliable picture from that. Twenty minutes, not two hours.
Step 3: Record five things for every answer
Keep a simple spreadsheet. One row per run, five columns:
Mentioned at all? Yes or no. This is your baseline visibility.
Position. First name in the list, buried in the middle, or a grudging "you could also consider"? Position in an AI answer works like position on a search results page. Most buyers never get past the first two or three names.
Accuracy. Is the description of your product actually right? Wrong pricing, discontinued features and outdated positioning are extremely common, and they're quietly costing you deals with buyers you'll never meet.
Sources. When the engine cites sources (Perplexity always does, the others sometimes), where is your mention coming from? Your own site, a review platform, a comparison article, a Reddit thread? This tells you what's driving your visibility, or what would fix its absence.
Who wins. Note which competitors appear when you don't. The brands that show up consistently are doing something citable that you aren't. That's your homework list.
Step 4: Score it honestly
You don't need anything fancy. Count the percentage of runs where you were mentioned, and note your average position when you were.
As a rough benchmark from the audits I run: brands with a healthy AI presence appear in 60% or more of category and competitor runs, usually in the top three names. If you're under 20%, AI assistants have effectively decided you don't exist in your category, and every buyer who starts their research in a chat window is starting it without you.
One number to hold in mind while you look at your score: AI chatbots are now the single most influential source for B2B vendor shortlists, ahead of review sites. This isn't a side channel any more. It's the front door.
Step 5: Repeat it monthly
AI answers shift constantly as models update and new content gets indexed. A score from January tells you little about April. Put a recurring 20-minute slot in your calendar, use the same questions with the same wording, and track the trend rather than any single result.
The trend is where the useful signal lives. Visibility that's climbing tells you your content and third-party mentions are landing. Visibility that's flat while a competitor's climbs tells you exactly where your next quarter of marketing effort should go.
What to do with what you find
If you're invisible, the fix usually isn't on your website. AI engines lean heavily on third-party signals: review platforms, comparison articles, community discussions and independent mentions. A site rewrite won't move the needle if nobody else on the internet is saying your name. (That's a full article in itself, and it's coming next.)
If you're visible but described wrongly, that's more urgent than it sounds. You can't correct an answer a prospect saw in a private chat. You can only fix the sources the answer was built from.
Or let me run it for you in 60 seconds
The manual version above works, and I'd genuinely encourage you to do it at least once, because seeing your competitor recommended in your place is more motivating than any statistic.
But if you'd rather skip the spreadsheet, I've built a free mini audit that queries the major AI engines about your company and sends you a visibility score in about a minute. No call, no email sequence ambush, just your score and what it means.
[Run your free AI visibility check →]
And if the score stings, that's what the full audit is for.
