Free Growth Assessment
Free Self-Check

AI Visibility Checker: A Free Self-Check for ChatGPT, Gemini and Perplexity

An AI visibility checker shows whether ChatGPT, Gemini and Perplexity name your company when buyers ask about your category. This free self-check gives you 12 prompts and a scoring sheet you can run yourself in an afternoon.

Last updated: November 21, 2026
12-Prompt Set
Category, comparison, problem and brand questions you can paste as written.
Fresh-Session Rules
Six setup rules so your own chat history doesn’t flatter the result.
0 to 3 Scoring Sheet
One repeatable number per AI tool, so next quarter’s run can be compared.
What the Score Means
Four bands, a cause table and a clear point to stop and call for help.
What It Is

What This AI Visibility Checker Is (and Isn’t)

Try it right now. Ask ChatGPT who the best provider of your service is in your market. Are you on the list?

An AI visibility checker is any method that tests whether AI answer tools find, name and correctly describe your company when buyers ask about your category, and this one is run by hand: you paste 12 prompts into ChatGPT, Gemini and Perplexity, log what comes back and score it on a sheet.

Don’t expect a scanner. It won’t watch your brand while you sleep. If you’d rather have a person trace every gap back to its cause, that’s the job of our AI visibility audit.

The thinking behind the prompts sits in our guide on how to measure AI search visibility. We’re not an agency. We’ve built, scaled and sold tech companies ourselves, so a score only matters to us if it moves pipeline.

Marketing director filling in a blank grid worksheet at a dark desk at dusk
Where It Breaks

Where a Free Check Misleads You

Here’s a pattern we see a lot. Someone on the team asks ChatGPT about the company, spots the name in the answer and calls it done.

But they asked by name, and a buyer never does that, because a buyer asks for a category first and only learns about you if the answer chooses to offer your name.

Then the other problems stack up. Settings carry over from earlier chats, the answer shifts on the second ask, and only one tool got tested.

Not ideal. A good score from a bad test is worse than no test at all, because it stops anyone from looking any further.

Six Ways a Quick Check Fools You
Each one makes the score look better, or worse, than it is.
Your own chat history and settings shape the answer, and a buyer’s won’t match yours
You asked once, and the next ask would have said something else
Every prompt had your company name in it, so the tool had nothing to decide
One tool was tested, and your buyer may use another
Nothing was written down, so there’s no baseline to compare against
A passing mention got counted as a recommendation
Before You Start

Set Up the Check Before You Ask Anything

Ten minutes of setup decides whether the result means anything at all, so follow these six rules and the numbers will be fair to you and to the competitors you’re being measured against.

Rule Why it matters How to do it
Start fresh Earlier chats and settings can color the answer Use a logged-out window or a clean browser profile, and open a new chat for every prompt
Fix the wording A reworded prompt is a different test Paste each prompt exactly as written, every time
Ask three times Answers can change between asks Run each prompt three times per tool, each in a new chat
Test more than one tool Buyers don’t all use the same one Use ChatGPT, Gemini and Perplexity, and add Claude if your buyers use it
Watch the competitors A score means little without a comparison Note which rivals are named in each answer
Write the date down The tools change, and so do the answers Log the date, the tool and the prompt for every run

No clean browser profile handy? A private window is a fair stand-in. Just open a fresh chat each time.

The Prompt Set

The 12 Prompts to Paste

Swap the brackets for your own service, market and company name, then paste. Each group of three tests something different, so don’t trim the list down to the prompts you expect to win.

Group Prompt to paste What it tells you
Category 1 Who are the best providers of [service] for a company like mine in [market]? Whether you’re on the shortlist at all
Category 2 What should I look for when choosing a [service] provider, and who is worth talking to? Whether the tool names providers while it advises
Category 3 I run a [size] [industry] company. Who would you recommend for [service]? Whether you fit the buyer you want to attract
Comparison 1 How does [brand] compare with [competitor] on price, scope and fit? How you’re framed next to a rival you pick
Comparison 2 What are the best alternatives to [competitor]? Whether you show up as the alternative
Comparison 3 Is [brand] or [competitor] better for a growing [industry] company? Which of you the tool leans toward, and why
Problem 1 How do I fix [problem your service solves], and who can help? Whether you’re named when a buyer starts from the pain
Problem 2 What does [service] cost for a small [industry] business? Whose pricing story gets quoted, yours or a rival’s
Problem 3 What are the warning signs that I need [service]? Whether your content shapes how the problem is described
Brand 1 What does [brand] do? Whether the tool knows you, and gets it right
Brand 2 Is [brand] any good, and what do customers say? Which reviews and profiles the answer leans on
Brand 3 Where is [brand] based, who runs it and what does it specialize in? Whether the basic facts about you are correct

Brand prompts flatter you. Give a tool your name and it has an easy job, but the buyers worth winning haven’t heard of you yet, which is why the first nine prompts deserve your closest attention once the results come in.

Two colleagues comparing three blank printed sheets at a standing table
How It Works

How to Run the AI Visibility Checker

Four moves, none of them hard. Clear an afternoon anyway: 12 prompts, three tools and three runs each adds up to 108 asks, and by about the fortieth you’ll be glad you wrote every answer down right away.

Got a colleague? One pastes, one records. That alone keeps the log honest.

The Four Moves

From Blank Sheet to Baseline

Write. Ask. Log. Score.

1
Step 1

Write Your Prompt Set

Start with the table above and swap every bracket for your details. Then drop in two or three prompts taken word for word from questions prospects asked on sales calls. After that, the wording is frozen, so a second run tests the same thing as the first.

2
Step 2

Ask Each Tool Three Times

Paste each prompt into ChatGPT, Gemini and Perplexity, opening a new chat every time. Resist editing it, asking a follow-up or arguing with the reply, because you’re taking a measurement here and not coaching the tool.

3
Step 3

Log Every Answer

For each run, write down the date, the tool, the prompt, whether you were named, how you were described, which sources were cited and which competitors showed up. A plain spreadsheet works fine.

4
Step 4

Score the Sheet

Give each prompt a score from 0 to 3 using the sheet below, then add up the scores for each tool. That’s your baseline. Save it, because the second run is where it gets useful.

The Scoring Sheet

The 0 to 3 Scoring Sheet

Score each prompt once per tool, using the result that showed up in at least two of your three runs. If all three runs came back different, take the lowest score. Be hard on yourself here.

Score What you saw In plain terms
0 Your company isn’t mentioned Missing
1 Mentioned, but the description is wrong, out of date or unclear Present, not trusted
2 Named correctly in the answer On the list
3 Named correctly, with a link or source that points to your own page Named and cited

Twelve prompts at a top score of 3 gives a maximum of 36 per tool. Here are the six columns to keep in your log for every run.

  • The date and the tool
  • The exact prompt and the run number, 1, 2 or 3
  • Whether you were named
  • How the answer described you
  • The sources or links the answer cited
  • The competitors named in the same answer

Tempted to skip the log? Don’t. It’s the only part of this whole exercise that lets you compare this run with the next one, and without that comparison a score is just a number somebody typed.

Reading the Result

What Your Score Means

Add up the 12 scores for each tool, then find the band. These bands are a rule of thumb we wrote for this checklist and not an industry benchmark, so use them to decide what to look at first and don’t quote them to anyone as a standard.

Score out of 36 Band What it usually means
0 to 9 Missing The tools rarely name you. Start with access, page content and how others describe you.
10 to 18 Patchy You’re named sometimes, usually on brand prompts. The category prompts are the gap.
19 to 27 Present You’re on the shortlist for some questions. The weak prompts show which pages to strengthen.
28 to 36 Strong The tools name you and often cite you. Keep the pages fresh and rerun every quarter.

If You See This, Look Here

The band tells you how bad it is. The pattern tells you where to look.

What you saw Likely cause Where to look first
Named on brand prompts only The tool knows you exist but has no reason to offer you Your service pages and the outside lists the answers quote
Competitors named, you aren’t The category answer was built without you The pages and listicles those answers cite
Described wrongly Old or conflicting facts about you Your site, your LinkedIn page and your directory profiles
Named but never cited Someone else’s page is the clearer source Whether your page states the answer in plain text
Low everywhere The bots may not be able to get in Your robots.txt and firewall rules

We don’t publish a citation rate for tech companies, because we don’t have one we can source. Your score compares you with yourself over time, not with a market average. For context, see our AI search visibility benchmarks guide.

The Free Technical Check

Check the Door: Can the AI Crawlers Reach You?

If your score is low everywhere, check this before anything else. Type your domain followed by /robots.txt into a browser and look for these four things.

  • A blanket Disallow rule that covers every bot
  • Named rules for OAI-SearchBot, GPTBot, ClaudeBot, Claude-SearchBot or PerplexityBot
  • Bot-blocking settings in your firewall or CDN
  • A rule somebody copied from a template years ago
Two settings

is what OpenAI lists for ChatGPT. Allow OAI-SearchBot to appear in search, and disallow GPTBot to opt out of training.

OpenAI crawler documentation

Three robots

is what Anthropic lists. Blocking Claude-SearchBot may reduce your visibility and accuracy in search results.

Anthropic support documentation

Allow it

is Perplexity’s advice for PerplexityBot, which links to sites in search results and isn’t used for model training.

Perplexity documentation

No extra step

is Google’s word on AI Overviews and AI Mode. Their visits show in Search Console under the Web search type.

Google Search Central

That’s each company’s own wording, not ours. Bot names and settings change, so read the current page before you edit anything.

Know When to Stop

When to Stop Self-Checking

A self-check answers one question. Are we named? It can’t tell you why not, and that part takes someone reading the cited sources, your pages and your robots.txt side by side until the pattern shows.

So here’s a simple rule. Run it once, run it again next quarter, and if the score hasn’t moved, stop guessing and get help.

When you want a person to trace each gap, that’s our AI visibility audit. Our AEO services page and SEO, AEO and GEO for tech companies show where the fixes land, and what AEO and GEO services cost explains what to expect.

Consultant and founder reviewing a bound report at a dark conference table

Is the Self-Check Enough?

The Self-Check Is Enough
  • You want a quick baseline before you decide anything
  • Someone on your team can own the fixes
  • Your score already lands in Present or Strong
  • You’re checking one service line or one market
Book the Audit Instead
  • You score Missing or Patchy and don’t know why
  • Competitors are named and you aren’t
  • The answers describe your company wrongly
  • You need findings tied to pages, not just a number
Common Questions

Questions About Checking AI Visibility

What is an AI visibility checker?

An AI visibility checker is a test of whether AI answer tools name, cite and correctly describe your company. Some are paid software that run the prompts for you. This one is manual, so you paste the prompts and score the answers yourself.

Are free AI visibility checkers accurate?

Accurate enough for a first look, not for a decision. Answers change between asks and between tools, so one run is a snapshot. Before you trust any score, find out how many prompts the tool ran and whether it repeated them.

How many prompts should I test?

Twelve is a workable first set. Add more from real sales calls as you go, because a bigger set steadies the score, but only if every prompt stays fixed from one run to the next and every single answer gets logged.

Do I need to be logged out to test?

Logged out is the safer choice. Your own chat history and settings can shape an answer in ways a new buyer won’t see. A private window or a clean browser profile works, and a fresh chat for each prompt stops earlier questions leaking in.

Why does ChatGPT answer differently each time?

Because these tools build each answer when you ask, and they can pull different sources each time. That’s why this checker asks every prompt three times and scores the result that appears at least twice.

Can I check Claude as well?

Claude is easy to add. Run the same 12 prompts in a new chat and score it the same way. Anthropic lists three robots, so look for ClaudeBot, Claude-User and Claude-SearchBot in your robots.txt too.

How often should I repeat the check?

Quarterly is a sensible rhythm, plus after any big change to your site. Keep the prompt set identical so the scores compare. A drop between runs is your cue to check robots.txt first, then your service pages.

Want Someone to Run It Properly?

Bring your score. We’ll go through what the AI tools say about you, whether the crawlers can reach your site and where to start. No deck. No hard sell.