Whether the machines advising your customers can see you
Your customer stopped asking Google. They ask AI — and AI answers with names.
We put 60 simulated customer questions to ChatGPT, Claude and Gemini about your exact industry, city and services. Then we show you plainly where you show up, where you get passed over, and who takes your place.
ChatGPT
OpenAI
detected
Claude
Anthropic
detected
Gemini
no signal
20 questions to each model
Google returns ten results. AI returns three names.
When a customer asks an assistant for a recommendation, they don't get a list to browse — they get a short answer with a few companies in it. Everything outside those lines is invisible.
Silence leaves no trace
Nobody tells you that you were skipped. No call, no enquiry, nothing in your analytics.
Your slot is already taken
If the model doesn't know you, it confidently recommends someone else — usually the same two or three names, over and over.
Wrong facts travel
An old address, a service you dropped, the wrong specialisation — once a model has learned it, it repeats it to every customer after that.
How we get to the answer
We read the business
We read your homepage plus your services and products pages, and use the company name as the leading signal. Out comes a profile: industry, city, services, customers, competitors.
We write 20 questions
We don't ask "do you know company X". We ask the way a real customer asks: who does this in this city, who would you recommend, who's best for this specific problem.
We ask all three models
The same 20 questions go to ChatGPT, Claude and Gemini separately — 60 answers, collected into one comparable report.
We read the answers
For every answer we record whether you were mentioned, at what position, in what tone, which facts were right, and which competitors showed up instead of you.
What the full report gives you
Not a score out of 100 and three lines of advice — the whole picture, with the charts that show exactly where the chain breaks.
- A five-axis radar
- Recognition, position, tone, factual accuracy and depth of knowledge — for all three models side by side.
- A map of all 60 questions
- A question-by-question grid: where you appear, where you vanish, and which kind of query loses you.
- The competitor ranking
- Which names come up most often in place of yours — and how many times each one does.
- Right and wrong facts
- What the models know correctly about you and what they repeat incorrectly, sentence by sentence.
- A prioritised plan
- Concrete steps ordered by expected impact — what to change first so the models learn you.
Question by question
Each cell is one customer question put to one model.
What we test — and what we don't
The report shows what the models already know from training, not the result of a live crawl of your site. That is exactly what happens when a customer opens an assistant and asks directly instead of going through Google. If a model uses browsing in a given conversation, answers there may differ. We say this plainly, because it's the difference between a useful report and a pretty picture.
One-time check
The full report for one business
- 60 answers from three models, read one by one
- Five charts showing where the chain breaks
- The competitors taking your place
- The facts the models get wrong about you
- A prioritised plan and a print-ready version
The overall score and per-model verdict are free. You only pay for the breakdown.
Questions we get
Do I have to pay just to see if I'm mentioned at all?+
No. The overall score and a one-sentence verdict for each of the three models are free. You only pay for the full breakdown — charts, competitors, facts and the plan.
Do you crawl my site live?+
We read your site to understand what you offer. But the models themselves answer from training — that's the scenario where a customer asks an assistant for a recommendation.
My site is thin. Is this still worth it?+
Yes, and that's exactly why we ask for the company name too. If the site is unreachable or nearly empty, the analysis starts from the name and industry and still produces a meaningful result.
Can I check several companies?+
Yes. One email can run as many businesses as you like — each gets its own report and its own link.
What do I actually do with the result?+
The report ends with concrete steps: structured data, a consistent name and address everywhere, content the models can read and remember.
Who is this for?+
Businesses that depend on being found — clinics, agencies, manufacturers, shops, consultants, local services.