Which brands do AI assistants actually recommend?
We ask ChatGPT, Claude, Perplexity and Gemini what to buy in a brand's own category, then count how often that brand is named.
The index has not been populated yet
16 brands are queued. The first results land shortly.
Run the audit on your own storeCould not be read
A store an assistant cannot read is a store an assistant cannot recommend. Because the request never reaches the origin, it does not appear in the store's own analytics either.
Both requests were answered by a Cloudflare managed challenge rather than the store. Sending a desktop Chrome user agent made no difference, so what is excluded is any client that cannot run a JavaScript challenge, not any client that declares itself a bot. The challenge page returns noindex,nofollow, so a crawler is turned away by a page the store never serves itself.
| Request | User agent sent | Response |
|---|---|---|
| Our crawler | Mozilla/5.0 (compatible; CresvaBot/1.0; +https://cresva.ai/bot) | 403 |
| Desktop Chrome | Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0 Safari/537.36 | 403 |
- Response signals
- cf-mitigated: challenge | server: cloudflare | <title>Just a moment...</title>
- Served to crawlers
- noindex,nofollow
How this is measured
Each brand is asked about in questions built from its own public catalogue, not from generic category templates, and those questions go to up to four assistants. A brand is counted as named only on a word boundary match against the raw answer. No model judges whether a mention happened, which is why two runs of an unchanged brand agree.
A score is only reported when at least two assistants answered at least four questions. Below that the run reports no score rather than a low one, because a quiet vendor and an absent brand are not the same finding.
The brand list is checked into the repository and changes through review. Appearing here is not an endorsement and there is no way to pay for a position. The full specification, including the failure modes this method has been wrong in before, is in the index methodology.