Open ChatGPT, Gemini and Perplexity, and type the exact question a buyer would ask, like 'who is the best [what you do]'. Note whether your name appears, where, and the reasons given. Screenshot it, since that record becomes your baseline for comparison later. That five-minute check is your starting line for People Engine Optimization, and it costs nothing but a little honesty about what you find.
Here's the thing. You can guess whether AI recommends you, or you can spend five minutes and actually know. Let's do the second one.
weekly ChatGPT users (Oct 2025)
of US adults have used ChatGPT
of under-30s have used it
The five-minute audit, step by step
You don't need a tool, a login, or a budget. You need the same apps your buyers already use. Open ChatGPT, Google's AI Mode, and Perplexity in three tabs. Then do this.
1. Ask the buyer's question, not yours
Type the thing a real client would type. Not "tell me about [your name]". That's cheating, the engine just repeats your bio. Ask the question that decides a sale: "who is the best fractional CFO for early-stage SaaS," or "who should I hire to fix my Shopify conversion rate." Phrase it like a busy human, superlatives and all.
2. Read the answer like a stranger would
Does your name show up at all? If it does, is it first, or buried in a polite list at the bottom? And crucially, what reasons does the engine give? It will often explain itself, and those reasons tell you which signal is carrying you or missing.
3. Screenshot everything
This is the part everyone skips and later regrets. Capture each answer, dated. This is your "before" picture. Without it, you can never prove anything moved, and you will move things.
Does the engine name you on the question that pays? Not "am I visible" in some vague way. A clean yes or no, on the exact question that would win you work.
Why the same question gets different answers on different engines
The first time someone runs this audit, they're often thrown by seeing ChatGPT name them while Perplexity draws a blank, or Gemini giving a completely different shortlist than either. That's not a bug, and it doesn't mean one tool is right and the others are broken. Each engine works from its own retrieval index, meaning its own particular slice and freshness of the web it has crawled and organized. Each also has its own training cutoff, the date up to which its core model learned general patterns, so an engine trained more recently may simply know about a newer article mentioning you that an older model never saw. On top of that, each engine weighs sources differently. One might lean more heavily on structured data and directories, another might weigh recent news coverage more, another might favor forum discussion. None of this is arbitrary, it's just different architecture producing different, equally real snapshots of how visible you currently are.
Reading a partial mention correctly
Not every audit result is a clean yes or no, and it's worth learning to read the middle ground correctly instead of over-reacting to it. A partial mention is when your name shows up in a longer list without being singled out, explained, or clearly recommended over the others, the engine acknowledges you exist in this space but hasn't yet decided you're the one to lead with. A full recommendation is when the engine names you specifically, gives reasons, and often puts you first or frames you as the standout choice. Both outcomes are useful data, but they call for different next moves. A partial mention usually means your Knowledge and Network signals exist but aren't yet dense or well-corroborated enough to win the tiebreak. No mention at all usually means there isn't yet enough attributable material connecting your name to the topic in the first place.
| Result | What it likely means | Where to focus next |
|---|---|---|
| No mention at all | No attributable body of work connected to your name yet | Publish foundational depth under your real name |
| Partial mention, buried in a list | You're on the radar but not yet the trusted lead answer | Earn third-party mentions and citations |
| Named first, with reasons given | Your signals are working, for now | Defend the position, keep publishing and monitoring |
Figure: how to read your audit result and where to point your next move, depending on what came back.
How to phrase your audit questions without biasing the test
It's tempting to type something like "why is [your name] the best choice for X," but that's a leading question, you're basically feeding the engine your desired answer and it will often oblige. A clean audit question sounds like something a stranger with money to spend would actually type, phrased around their problem, not your name. Keep three rules in mind: use the buyer's language and context, not your own branded terms, avoid superlatives that point at you specifically, and keep the exact phrasing identical every time you re-run the check so that a change in results reflects a real shift in visibility rather than a change in how you asked. If you want more example phrasing pulled directly from real buyer behavior, the prompts buyers actually type into AI is the natural next read.
A hypothetical: running the audit as a wedding photographer
Imagine a wedding photographer named Sanya, based in a mid-sized city, trying to figure out where she actually stands. Instead of typing "tell me about Sanya's photography," which would just have the engine repeat whatever bio it can find, she types the question a real engaged couple would ask: "who's a good wedding photographer in my city for a documentary style wedding." On ChatGPT, she isn't mentioned at all. On Perplexity, which leans more heavily on recent web pages and reviews, she shows up third in a list of five, with a short description pulled from her own site. On Google's AI Mode, nothing. That spread isn't a contradiction, it's simply three different systems drawing on three different slices of the web, at three different moments in time. Sanya's actual next move isn't to panic about ChatGPT, it's to notice that Perplexity found something real to work with (her site, apparently) and ask what would make that mention stronger and more consistent, and what would need to exist for the other two engines to find something similar.
What your results actually mean
Three outcomes, three meanings. If you're unmentioned, the machine doesn't know you exist in that context yet. That's not failure, it's a blank page. If you're mentioned but buried, you're on the radar but not trusted enough to lead, usually a network problem. If you're named first, congratulations, now the job is to defend it, because someone else is reading this same guide. If a specific competitor keeps beating you to the answer, it's worth diagnosing that gap directly in why AI recommends your competitor, not you.
Run it across engines, not just one
ChatGPT might name you and Perplexity might not. That's normal, they weigh sources differently. Checking all of them shows you where you're strong and where you're invisible, which is far more useful than one lucky screenshot you show your friends. If you want a fuller side-by-side breakdown of how these tools actually differ in behavior, Perplexity versus ChatGPT versus Gemini covers it in more depth than this audit alone can.
Common mistakes that quietly ruin the audit
A few habits sneak in and make the results less trustworthy than people realize. The biggest one is switching up the wording each time you check, which makes it impossible to tell whether a changed answer reflects real progress or just a differently phrased question. Another is running the audit only once and treating that single result as a permanent verdict, when in reality these systems can and do give slightly different answers on different days even with the identical question, simply due to normal variation in how they generate responses. A third is only checking the engine that happens to already be favorable to you and ignoring the ones where you're invisible, which flatters your ego but tells you nothing useful. And a fourth is forgetting to note down the actual reasoning the engine gave for its answer, which is often the most useful diagnostic detail in the whole exercise, since it tells you exactly which signal, published depth, third-party mentions, or consistent identity, is doing the work or falling short.
Building the monthly tracking habit
Then do it again next month. A single audit is a snapshot. The real value is the movie. Re-run the exact same questions monthly and watch the trend. Keep a simple log, the date, the exact question, which engine, whether you appeared, where, and what reason was given, so you can compare properly instead of relying on memory. Set a recurring reminder rather than trusting yourself to remember, this is the kind of check that quietly slides once the initial excitement fades. Over several months, this log becomes the clearest evidence you have of whether your published work and earned mentions are actually landing, and it's a far more honest measure of progress than any single screenshot. If you're curious what a realistic pace of change looks like once you start doing the work, how long until AI knows you lays out a fair timeline. Resist the urge to skip a month just because you didn't publish anything new, a flat or unchanged answer is still useful data, it tells you the current record is stable rather than slipping. That loop, measure, act, re-run, is the whole discipline. The audit isn't a one-time reveal, it's the scoreboard you'll be steering by from now on, and like any scoreboard, it only tells you something useful once you've checked it more than once.
Questions people ask
Which AI should I check first? +
What if I'm not mentioned anywhere? +
Is one screenshot enough proof? +
Why does ChatGPT name me but Perplexity doesn't? +
What counts as a partial mention versus a full recommendation? +
How do I phrase audit questions without biasing the test? +
Sources
Curious what AI says about you?
Start with a check-up. We'll show you the exact words the engines return about your name, then map the fastest signal to move.
Say my name →