In short. Cohere states that it uses no Cohere training crawlers and presents Coherebot as a conditional robots.txt example.

Definition: Cohere crawler policy is Cohere’s published position on automated web collection for model training. As of 17 October 2026, Cohere states that it uses no Cohere bots or user agents to crawl or scrape web content for foundation-model training.

Key facts

Cohere’s Web Crawlers documentation lists “N/A” for bot name and description, and presents Coherebot in a conditional blocking example.
The IETF’s Robots Exclusion Protocol, RFC 9309, defines how crawlers match their names to groups of website access rules.
Google’s robots.txt creation guide specifies that a file applies to its protocol, host and port.

What does Cohere’s statement cover?

The Cohere documentation addresses crawling or scraping web content to train foundation models. Its stated policy requires crawlers to be designed to respect robots.txt, the file through which websites publish crawler access rules.

Cohere says it will identify future training bots or user agents in the page’s table and provide blocking information. The example therefore describes a prospective control; the table identifies no active training crawler.

What does the Coherebot example mean?

Cohere’s conditional example contains these lines:

User-agent: Coherebot
Disallow: /

A product token is the crawler name used to select rules. Under RFC 9309, User-agent selects a group for a matching token, and Disallow identifies paths excluded from crawling. The slash covers paths beginning at the site root. Matching crawler names is case insensitive.

The presence of this example establishes neither an operating bot nor its observed request identity. Cohere says that blocking IP addresses can prevent access to the robots.txt file and make a persistent opt-out unreliable.

How can a website check the rule’s scope?

The robots.txt creation guide places the file at the root of the relevant host. Rules at https://example.com/robots.txt apply to that host and protocol; a subdomain needs its own applicable file.

RFC 9309 requires the file to use UTF-8 encoding and the text/plain media type. It also requires crawlers to combine groups matching the same token. When no specific group matches, the wildcard group, User-agent: *, applies if present. Reviewing the complete file therefore matters when interpreting a named rule.

FAQ

Should I treat Coherebot as an active training bot?

Cohere’s documentation presents that name as an example for possible future crawling. You can follow its changelog or RSS feed for policy updates.

Does a robots.txt rule secure private content?

RFC 9309 describes robots.txt as crawler instructions and says these rules provide no access authorization. You need an appropriate security measure to protect private content.

Sources

Check how a page reads to answer engines. Open the GEO score analyzer