Hundreds of millions of people now ask ChatGPT and Google’s AI who to hire, where to eat, who to trust. Those systems check a business before naming it, and a business that fails the check doesn’t get an error message or a warning letter. It gets silently replaced by a competitor who passes, in an answer you’ll never see.
Which makes this the strangest blind spot in small-business marketing: you can know your Google ranking to the decimal and have zero idea what the AI layer says about you. The good news: checking takes five minutes and costs nothing, because you can interrogate the machines directly. Five questions, asked the way real customers ask, read the way an auditor reads. Here’s the test, what each answer means, and the fix behind each failure.
The setup (30 seconds, matters more than it looks)
Open a fresh or temporary chat, your history contaminates results, the AI plays nice when it knows you’re the owner. Run the five questions in ChatGPT first, then repeat in Google’s AI Mode (the tab next to All/Images, or google.com/ai). Two systems, because they read different evidence and fail differently. Screenshot everything.
The five questions
Test 1: the existence check
- Ask: “What do you know about [business name] in [city]? Search the web if needed.”
- Pass: it describes the right business, current facts. Partial: it knows you but recites old prices, dead addresses, wrong hours. Fail: “I couldn’t find information about this business,” the AI-era version of not existing
Test 2: the recommendation check (the money one)
- Ask what a customer asks: “Who’s a good [service] in [city]?” then once more with a constraint: “…good with [urgent jobs / fair prices / old buildings]”
- Pass: you’re named, with a reason attached. Fail: competitors get named and reasoned about while you’re absent. Note WHO gets named and WHAT reason each gets; those reasons are the evidence beating yours
Test 3: the fact audit
- Ask: “What are [business]’s hours, location and typical prices?”
- Wrong facts are worse than no facts: a customer told you close at 5 doesn’t call at 6. Each wrong answer has a stale public source behind it, findable and fixable
Test 4: the reviews check
- Ask: “What do customers say about [business]?”
- Watch the split: Google’s AI usually quotes your Google reviews; ChatGPT often can’t see them and leans on whatever’s on the open web, the wall explained in the reviews-visibility guide. A thin ChatGPT answer here with 100 Google reviews is normal, and fixable
Test 5: the access check
- Ask: “Open [yourdomain.com] and tell me what this business does.”
- Pass: it reads and summarizes. Fail: “I can’t access this site,” meaning a blocker (often a Cloudflare default or security plugin) is turning AI away at the door, diagnosis in the crawler guide
Reading your scorecard
5/5: you’re in the visible minority; move to offense (the “because” game from the near-me playbook). Passed 1, failed 2: the commonest result, you exist but aren’t chosen; the gap is evidence: review sentences, liftable pages, a stated specialty. Failed 1 or 5: foundation problems, verification signals or blocked crawlers, start at the existence guide. Wrong facts anywhere: origin-hunt per the wrong-info method. Every failure maps to a fix; none of them is “wait and hope.”
Make it monthly, three lines in a note
Answers rebuild constantly: models update, competitors publish, your fixes land. The owners who win this layer re-run the five questions monthly and log three things: named or not, the reason given, facts right or wrong. Five minutes, and over a quarter the log turns into the most honest marketing report you own, it shows whether the most influential referrers on earth are selling you or your competitor. The same habit, automated and scored across eight signals, is exactly what our checker does, which brings us to the shortcut below.
Common questions
How do I check my business’s AI visibility?
Interrogate the AIs directly in a fresh chat: ask what they know about your business, who they’d recommend for your service in your city, your hours and prices, what customers say, and whether they can open your website. Run it in ChatGPT and Google’s AI Mode, and screenshot the answers.
Why use a fresh or temporary chat for the test?
Chat history biases answers: an AI that knows you’re the owner describes you more generously than it would for a stranger. Fresh chats show what an actual customer hears.
What does it mean if ChatGPT can’t find my business?
Your public evidence is too thin or inaccessible for verification: weak or inconsistent listings, no crawlable facts, or blocked AI access. It’s the AI-era version of not existing, and it has a standard repair path.
The AI knows me but recommends competitors. Why?
Recommendations need justifiable evidence: review sentences describing specifics, pages stating prices and specialties, a reason the AI can write down. Being findable is step one; being chooseable is an evidence upgrade.
Why does Google’s AI describe me better than ChatGPT?
They read different evidence: Google’s AI sees your Business Profile and review text natively, ChatGPT sees the open web. A Google-only reputation produces exactly that split, and republishing evidence on the open web closes it.
How often should I run this test?
Monthly, same five questions, three logged lines: named or not, the reason given, facts right or wrong. Answers change as models update and competitors publish, so a quarterly-or-never habit misses the plot.
Is there a faster way than asking manually?
Yes: an AI visibility checker automates the audit, scoring the signals AI systems verify, crawlability, facts, structure, reviews, and returning a fix list. The manual test stays valuable for hearing the actual answers customers hear.