Does My Business Show Up in ChatGPT? How to Check for Free

To check whether your business shows up in ChatGPT, ask it the questions a real customer would ask — several ways, with search enabled — and record whether you’re named, where you rank, and which sources get cited. One answer proves nothing; a pattern across many answers is your actual visibility.

This page gives you the full manual method for free: the exact prompts, why sampling matters, and what to write down. Then it covers the faster version — our free scan — and what a full 100+ answer audit adds if you want the complete picture.

Why this is worth checking at all

Real customers are already asking. 29% of adults now initiate daily searches via generative AI summaries, and among Gen Z adults, 82% have used AI chatbots — your youngest customers most of all. To keep it honest: about 84% of US consumers still use Google daily for local business searches, and only about 24% prefer AI chat for local queries. But the distinction is collapsing, because Google’s own answers are now AI-composed too.

So the question isn’t whether AI answers matter yet. It’s whether, when they’re given in your market, you’re in them. Most owners have never checked.

The manual method: ask like a customer

Open ChatGPT with search enabled. Don’t ask about your business by name first — customers who already know your name aren’t the ones you’re losing. Ask the way a stranger with a need would ask. Cover four intents:

Discovery — the open-ended scan:

• “plumbers in Frisco, TX”
• “mortgage brokers near downtown Boise”

Recommendation — the direct ask:

• “Who’s the best HVAC company in Tucson?”
• “Can you recommend a good family dentist in Carmel, Indiana?”

Comparison — the shortlist:

• “Compare the top real estate agents in Bend, Oregon for first-time buyers”

Task-specific — the urgent, high-intent version:

• “My water heater is leaking and I need someone in Plano today — who should I call?”
• “I need a realtor who knows the school districts in Frisco — who do people use?”

Swap in your own category and city. Then — this is the part everyone skips — ask each question again in two or three different phrasings. “Best plumber in Mesa,” “who should I call for a burst pipe in Mesa,” and “top-rated Mesa plumbing companies” are the same customer need, and they can produce different answers.

Finally, ask about your business by name: “Tell me about [Business Name] in [City].” You’re checking for wrong or stale data — old address, wrong hours, services you don’t offer, or “permanently closed.”

Why one answer is an anecdote

If you ask once, get named, and celebrate — or ask once, get skipped, and panic — you’ve measured noise.

Three sources of variance stack up. Models sample their wording, so the same prompt can produce different answers minutes apart. Small phrasing changes pull different retrieved pages. And the retrieved pages themselves shift as the web updates. Ask ten times and you might be named six — that’s not a glitch, that’s the actual shape of your visibility.

There’s a diagnostic upside to the variance, too. If you’re named in some phrasings but not others, the answers that skip you tell you which retrieved sources are missing you. If you’re named by one AI system but never by another, the systems are leaning on different sources — and the gap points at specific pages. Consistent absence everywhere is a different, more fundamental problem than patchy presence.

This is why visibility is a rate, not a yes/no. It’s also why our audit battery exists: 20 localized prompts — four intents in multiple phrasings — run against 6 AI systems, for 100+ timestamped answers per audit. Each response is stored raw and quoted verbatim. Not because 100 is a magic number, but because one is meaningless and a dozen is a start.

For your manual check, a realistic target: 8–12 questions across two AI systems, each asked once. That’s an afternoon coffee’s worth of work and enough to see a pattern.

What to record

Write down more than “yes/no.” A simple spreadsheet with one row per answer:

Field	What to capture
Date and time	When you asked (answers drift over time)
System	ChatGPT, Perplexity, Gemini, Google’s AI answers
Exact prompt	Verbatim — you’ll re-run these next month
Named?	Yes / no
Position	First mention, mid-list, afterthought
What it said	Verbatim quote, especially anything wrong
Cited sources	Every domain the answer linked or referenced

The cited-sources column quietly becomes your most valuable asset. Across a dozen answers, the same few domains will keep appearing. That’s the start of a citation-source inventory — the list of pages that actually determine your AI visibility in your market. If you’re absent or outdated on them, you know exactly why ChatGPT isn’t recommending you.

Two rules from our own methodology worth stealing: record answers verbatim — never paraphrase, because “close enough” hides the errors that matter — and timestamp everything, because you’ll want to compare against next month.

The fast version: the free scan

If you want a first read without an afternoon of prompting, that’s what our free preliminary scan is for. It runs 3 real customer questions across 2 AI models plus Google, takes about 30 seconds, and emails you the results — whether you were named, what was said, and what got cited.

It’s deliberately small: a preliminary read, not a diagnosis. Think of it as the same manual method above, automated and shrunk to its minimum useful size. If it shows you’re consistently visible, good — check again in a quarter. If it shows gaps, you’ll want a bigger sample before deciding what to fix first.

What the full audit adds

The full audit is the manual method taken to its logical end. The differences are sample size and scoring:

• 100+ answers instead of a handful. 20 localized prompts across 6 AI systems (GPT-, Claude-, Gemini-, DeepSeek-, Qwen-class models plus Google’s answer layer), each stored raw with a UTC timestamp.
• A Visibility Score (0–100). Weighted: mention rate 50%, prominence 20%, sentiment/accuracy 15%, source presence 15%. What counts as a good score depends on your market — we’ve written up what a good AI visibility score looks like separately.
• Share of voice. Not just whether you’re named, but how often relative to the competitors who are — the number that stings, and the one that moves.
• A complete citation-source inventory. Every domain the answers drew on in your market, so the fix list writes itself.
• Monthly re-audits with the identical battery, so month-over-month change is real signal, not phrasing luck. The full methodology is on how it works.

Nobody can guarantee AI mentions — we can’t and neither can anyone else. What measurement gives you is the honest baseline and proof of whether the work is working.

What to do this week

1. Pick 8 customer questions for your category and city using the four intents above — at least two phrasings for your most important question.
2. Ask them in ChatGPT with search enabled. Save every answer verbatim, with date and time.
3. Repeat the same 8 in one more system (Perplexity or Gemini) for a second sample.
4. Ask each system about your business by name and check every fact it states against reality.
5. Log everything in the seven-column format above, and highlight every cited domain.
6. Tally your mention rate: answers naming you ÷ total answers. That’s your baseline number.
7. Calendar the same check for the same day next month, same prompts — a number without a trend is just trivia.

Or skip the setup and get your first read in about 30 seconds with the free AI visibility scan.