Cohere AI

Cohere AI is the directory entry for requests containing the token cohere-ai. The entry classifies it as an AI training crawler and retains a description that says it collects public web text for Cohere's models. Cohere's current operator documentation does not confirm that identity.

In its web crawler policy, Cohere says it does not currently use Cohere bots or user agents to crawl or scrape web content for generative AI foundation-model training. Its table of active crawler names contains no entries. The policy says Cohere would identify any future training crawlers and publish blocking instructions.

That conflict changes how a log match should be interpreted. A cohere-ai header is not current evidence that Cohere fetched the page or placed its content in a training dataset. The available sources do not establish whether the token is obsolete, used by another client, or simply spoofed, so none of those explanations should be asserted without request-level evidence.

Cohere models can be connected to search tools by developers, but Cohere's documentation has the external tool perform the search. Traffic from such a setup would carry the identity chosen by that tool or application. The cohere-ai token is not documented as Cohere's public AI search agent, and allowing it has no verified effect on citations or search visibility.

This entry marks robots.txt support as true. Cohere's current policy says any future crawler must be designed to respect robots.txt, but its example uses the hypothetical token Coherebot, not cohere-ai. Robots behavior for requests using this page's token is therefore unconfirmed despite the retained metadata value.

There is no Cloudflare bot-directory record, Web Bot Auth signature directory, or verified IP status for cohere-ai in the supplied facts. A site can classify or block the literal token, but should not grant access or attribute the request to Cohere on that basis alone.