Where every number comes from.

What we ask

Five questions your customers ask, written from your website: the best tools in your category, what your company is, you against your closest rival, alternatives to that rival, and a question about the problem that names no brand.

The free check asks Gemini each question once. The Deep report asks each question twice on ChatGPT, Gemini, Claude and Perplexity, so one unusual answer cannot move the result.

The score out of 100

Each answer earns up to 100 points. The report shows the average, and a separate score for each AI tool.

Proof

Every number links to the answers behind it. You can open the exact answer each AI gave, with its date, and see where your company is named.

What is asked

Up to 25 buyer questions (5 during the free trial), written from your website and confirmed by you, are sent to ChatGPT, Claude, Gemini and Perplexity every week, worded the same way every time. Each question is asked more than once on each AI.

The headline number

Week by week, the dashboard shows one simple number: in how many answers to discovery questions did AI name you? Discovery questions are the ones a buyer asks before knowing any name. Questions that already contain your name are shown separately, because naming you there proves nothing.

Position, tone and whether your website is quoted are shown as separate tiles, so you can see exactly what moved.

The range, and the Early read label

The number is shown with a range, for example “named in 8 of 24 answers (33%), likely between 18% and 52%”. The range is the honest margin from the number of answers. Until there are at least 60 answers, or while on trial, the result is labelled Early read.

What counts as a real change

AI answers vary. A change between two checks is called real only when it is bigger than the margin of both checks combined (a standard two-sample test at 95%). Anything smaller is shown as Unclear, never as a rise or a drop. A change is only compared when both checks used the same questions.

Weeks, months, quarters and years

Each comparison uses the average of the complete weekly checks in the period. Months compare calendar months, quarters compare calendar quarters, and years compare the same month a year apart, so seasons match. A period stays locked until enough checks exist, so nothing is guessed.

Check it yourself

Every question has links to ask the same thing in ChatGPT, Perplexity, Claude and Gemini, and buttons to tell us whether your result matches ours. We also spot-check the AI tools ourselves. One whose results often do not match is marked “indicative only”.

Pages written for you

Nothing in a page we write rests on a claim without a source. Facts about your company are approved by you. Facts about rivals come from their own public pages, saved with the date, and price facts expire after 30 days. A separate checker compares the finished draft with those sources before it is delivered.

Honest limits

It shows what AI answered. It does not predict rankings, traffic, leads or revenue, and it cannot prove why an answer changed. Suggested next steps are based on gaps visible in the answers. Results from publishing a page usually take weeks to show, and nothing is guaranteed.