This is what AI Radar shows you every week. A sample account: Fernway Coffee (a fictional brand), 14 weeks of scans across ChatGPT, Perplexity, Claude, Gemini and Google AI.
Proof page: what the engine said, who it cited, and the week-by-week record for this one answer.
The answer did not name atlascoffeeclub.com in a sentence we can quote. It linked to it as a source.
Claude left Fernway Coffee out in 3 of 3 runs this week and cited "atlascoffeeclub.com" instead. That page is the one to beat, and it is a page you can beat: it is a competitor's own site, not a community or an authority.
Claude reads through the Brave index and prefers a neutral, sourced register with a named author. Superlatives count against you here.
Not checked yet. Opens the page the way each AI tool’s bot does and reads robots.txt. About 10 seconds, free.
Scores both pages on 8 things that make a page easy for AI to quote, and shows check by check what theirs has that yours does not. Takes about 90 seconds. Up to 8 new pages a week per workspace.
| Week | Linked you | First link |
|---|---|---|
| 27 Sept | 0 of 3 runs | atlascoffeeclub.com/ |
| 4 Oct | 0 of 3 runs | atlascoffeeclub.com/ |
| 11 Oct | 0 of 3 runs | atlascoffeeclub.com/ |
| 18 Oct | 0 of 3 runs | atlascoffeeclub.com/ |
| 25 Oct | 0 of 3 runs | atlascoffeeclub.com/ |
| 1 Nov | 0 of 3 runs | atlascoffeeclub.com/ |
| 8 Nov | 0 of 3 runs | atlascoffeeclub.com/ |
| 15 Nov | 0 of 3 runs | atlascoffeeclub.com/ |
Everything quoted is what the AI tool returned in the scan. The fix is Radar's read of what usually wins this kind of answer, not a claim about why the AI tool picked that page. Won is decided only by the weekly scan.