How to Check If ChatGPT and Perplexity Recommend You

"We rank number one on Google for our category." The founder said it the way people do when they think the conversation is over. So I turned my laptop around, opened ChatGPT, and typed the exact sentence one of his customers would type before hiring anyone: best commercial HVAC company in his city, and who to call first. The model wrote four confident paragraphs. It named three competitors. It did not name him once.

He had never done this. Not because he was careless, but because nothing had ever told him to. Google gives you Search Console: impressions, average position, the query report, a dashboard that says exactly where you sit for every term that matters. The models give you nothing. No console, no share-of-voice report, no weekly email flagging that you slipped. You are being recommended, or quietly skipped, in front of buyers every day, and the surface that would show you that does not exist yet.

So you build the read yourself. It is not hard. It is a discipline nobody has assigned you, because the tools that usually nag you about your visibility have no window into a chat that answers a buyer and closes before you ever see it.

How do I check if ChatGPT recommends my business?

Open ChatGPT, Perplexity, and Gemini, and type the actual questions your buyers ask before they buy, phrased the way a real person phrases them, then log which businesses each model names. That is the whole test. You are not checking whether your website ranks. You are checking whether the machine that now answers before the links load says your name out loud.

The queries that matter are not the ones you would type. You know your own brand name, so searching it just shows you yourself, which proves nothing. Your buyer does not know your name yet. Your buyer types the category and the intent: who does the best work in this town, what should I look for, is that competitor any good, who is cheaper than the place my neighbor used. Those are the prompts that decide whether you exist in the answer or not, and they are the ones you have never run because you already know who you are.

Run each one cold, in a fresh session with no memory of your earlier questions. Log the response, the businesses named, the order they land in, and whether you show up at all. Then, and this is the part that turns a party trick into an instrument, do it again next month. The answer is not fixed. The model updates, your competitors publish, your review count moves, and your share of the answer drifts without warning.

The prompt set is the instrument

A one-off search is a curiosity. A prompt set you run the same way every month is a measurement. Build it once and it becomes the closest thing to a Search Console the models will give you for a while.

• Write ten to fifteen prompts per category. Mix the obvious ("best med spa near this suburb") with the comparison ("is that clinic or this one better for a first visit") and the objection ("who is more affordable than the expensive option everyone names"). These are the shapes a real buyer uses when the model is doing the shortlisting for them.
• Run every prompt across all three engines. ChatGPT, Perplexity, and Gemini pull from different places and reach different answers. Being named by one and skipped by two is a real and common result, and you only see it if you check all three.
• Log names, not vibes. Record the exact businesses each model lists and where you fall in the order. "It felt like it mentioned us" is not data. A sheet with names and positions is.
• Watch the sources. Perplexity shows its citations outright, and the others will sometimes name theirs. The pages a model leans on to build its answer are the pages you need to be inside of, which is the whole game of getting quoted instead of just ranked.
• Rerun the identical set monthly. Same prompts, same engines, same day. Consistency is what converts a screenshot into a trend line.

The uncomfortable part shows up once you actually run it against real local categories. Ask the models for the best med spa near a specific suburb and they do not read your homepage and admire it. They assemble an answer out of directories, review counts, and the handful of pages that describe your category in plain, structured language a machine can parse. The business with sixty surfaced reviews and a clear services page gets named. The business with a gorgeous site, thin review presence, and copy written for a human who is already sold gets skipped. The model is not judging your taste. It is pattern-matching on the signals it can read, and a slow, review-light, jargon-heavy presence is close to invisible to it.

Ranking number one is a shrinking flex when the model answers before the links load.

Isn't this the same as measuring AI referrals?

No, and the two answer different questions. Measuring AI referrals asks whether ChatGPT is already sending you clicks and customers, which you read in your analytics after the fact. This check asks the question upstream of that: whether the models name you at all, before anyone clicks anything. You can be invisible in the answer and still catch the occasional referral, and you can be named constantly and never bother to measure the traffic. Two instruments, same dashboard, different jobs.

The distinction is worth holding because the fixes differ. Whether ChatGPT is sending you customers is the downstream number you track in analytics. Getting quoted and cited is the mechanism that earns you the mention in the first place. This post is the audit that sits between them: a repeatable read on whether the mechanism is working, run on your own category, so you learn whether you have a visibility problem before you spend a quarter fixing one you might not have. Skip the read and you are guessing, and guessing about your own presence is how a competitor cements the top of the answer while you are busy assuming you own it.

If you want the automated first pass, our free Pre-Flight Check reads the machine-facing signals the models lean on: site health, structured data, local presence, and the review surface that decides whether you are legible to an answer engine at all. It will not type prompts into ChatGPT for you, but it tells you fast whether the underlying signals exist for a model to find. Most invisibility starts there, in the signals, not in the wording of any single page.

How often should you run the check?

Monthly, logged as a tracked metric, not a one-time gut check. AI answers are not static rankings; they shift as models retrain, as competitors publish, and as your review counts and citations change underneath you. A single check tells you where you stand today. A monthly log tells you the direction you are moving, which is the only reading that lets you act before someone else owns the top of the answer for good.

Treat it like any owned instrument. Same prompt set, same three engines, same day of the month, results in a sheet you keep. Over a few months the pattern gets legible: you are gaining names, holding steady, or slipping. That trend line is your share of voice in the one place there is no official console to hand it to you. It is the difference between managing your AI visibility and finding out about it the day a customer mentions they asked ChatGPT and it recommended the shop across town.

What you do with the reading is the same discipline we argue for everywhere else. The review engine that makes a local business legible to an answer model is not rented; it is built and owned, the way the Skin & Self engagement produced 4.9 stars across 757 reviews from an automated system rather than a bought batch that evaporates. The structured, category-clear pages the models quote are the same pages that win the local map pack on purpose. All of it is infrastructure you own instead of growth you rent, which is the only kind that keeps compounding once the platform changes its mind about how it answers.

The gap between being named and being skipped is not luck, and it is not permanent. It is a set of signals you can measure this week and move over the next quarter. If your own test comes back with three competitors and none of you, that is a visibility problem with a known fix, and it is worth a conversation before the answer hardens around someone else. Book a call and we will run the read with you, on your real category, and tell you honestly whether you have a gap worth closing.