Updated
AI Visibility Report
Whether ChatGPT, Claude, Gemini, Perplexity and six more name you, scored and retestable.
When a buyer asks an assistant about what you sell, do you get named, or does a competitor?
Businesses who suspect buyers are asking an assistant about their category before they ever reach a search box.
Do the assistants name you?
Your buyers ask them before they ask you. The report answers it with a number.
Share of your buyer questions where you get named, weighted by where in the answer you appear. First mention counts more than a passing one at the bottom.
-
ChatGPT
Not named
-
Perplexity
Named 3rd
-
Claude
Not named
-
Gemini
Named 1st
-
Copilot
Not named
-
Google AI Overviews
Cited
-
Grok
-
DeepSeek
-
Mistral
-
Meta AI
An illustration of the output format. Your engines, your questions, your number.
- A number you can retest. Same questions, same engines, so you can tell whether anything you change actually moved it.
- The competitors named instead of you, per question, so you know who you are actually losing to inside an answer.
- The sources those answers were built from. Usually directories and comparison pages rather than anyone's own website, which tells you where the work is.
- Whether the assistants can reach your site at all. A quarter of the firms in my last audit blocked them without knowing.
You can, and you should, once. What you cannot do in an afternoon is ask ten engines the same buyer questions, repeat the runs across several days because single answers vary, score the results the same way each time, trace which sources each answer was built from, and check whether your site blocks the crawlers behind them. That collection and repetition work is what the price buys. The judgement on what to do about it is on top.
Four sample pages, so you know what arrives.
The structure, depth and evidence style below are what you get. Every name and number on these pages is example data, blacked out the way a client copy would be redacted.
is named in around a third of the answers a buyer would see. Two competitors appear more often, and one engine cannot reach the site at all.
- 10AI engines tested
- 25buyer questions
- 3runs per question, days apart
- 750answers scored
Page 1 of 14. The score is retestable: same questions, same engines, same scoring.
| Engine | Names you | What it says | Named instead |
|---|---|---|---|
| ChatGPT | yes | Second of four firms listed, description matches the service | |
| Google AI Overviews | no | Answer built from two directories, no firms named | |
| Perplexity | yes | Named once in three runs, cited from a comparison page | |
| Gemini | no | Describes the service wrongly, using an out of date page | |
| Claude | no | Cannot fetch the site: crawler blocked at the firewall | |
| Copilot | yes | Named with a price that has not been correct since 2024 |
Plus four more engines in the full report, each with the transcript behind the verdict.
Where the answers come from matters more than who is in them: in this example, 62 percent of citations were directories and comparison pages, not anyone's own website. That is where the work is.
Times named across the 25 questions. Your row is highlighted; competitor names are redacted in this sample.
- Unblock the AI crawlers at the firewall. One engine cannot read the site at all. Until it can, nothing else on this list helps with it.
- Get onto the two directories the answers are built from. Both list competitors and not . Named, with submission links.
- Correct the out of date page Gemini is quoting.
Ordered by impact. Each action names the evidence page it comes from, so you can hand the list to anyone.
In the report
- A visibility score across ten AI engines, and the buyer questions it was measured on
- Every prompt run three times across several days, because one answer is a sample and not a finding
- Which competitors get named instead of you, per question, ranked by how often
- Which sources those answers were assembled from, which is where the work usually is
- How each engine describes you, and whether that matches what you actually sell
- AI crawler access check across GPTBot, ClaudeBot, PerplexityBot and Google-Extended
- A repeatable baseline, so a retest in three months measures the same thing
Not in the report
- Fixing what it finds. That is quoted separately once you know the size of the gap.
- Writing content or building citations.
- Ongoing tracking. This is a point-in-time baseline, not a subscription.
Saying what a thing does not include is the fastest way to stop you buying the wrong one.
Three things, then I start.
You get asked for these on the confirmation page the moment you pay. Nothing else is needed and there is no call to book.
- Three or four questions a real buyer would type before shortlisting a firm like yours
- The two or three competitors you actually lose deals to
- Your domain, and the country your buyers are in
Still not sure it is the right one?
Describe what happened and when. If this report is wrong for you, I would rather say so now than take £295 for it.
Ask first