MistralAI User: Mistral reading your page for a user
The MistralAI user agent, MistralAI-User, handles user actions in Vibe, Mistral's assistant. When someone asks a question, Vibe may visit a web page to help answer and then include a link to that source.
Mistral is specific about what it is not. The agent does not crawl the web automatically, and Mistral says it does not collect content for generative AI training.
Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; MistralAI-User/1.0; +https://docs.mistral.ai/robots)
- Operator
- Mistral AI
- Type
- Live fetch for a user
- robots.txt token
MistralAI-User- Obeys robots.txt
- Yes, by the operator's documentation. Mistral says the token governs which sites user requests can reach
- Official docs
- docs.mistral.ai
Three Mistral agents, three tokens
Mistral documents a small family of agents on one page. Each has its own token, so you can treat them one by one.
| Agent | Job | Training |
|---|---|---|
| MistralAI-User | Fetches a page for a question in Vibe | Not used for training |
| MistralAI-Index | Crawls to index pages for Mistral search | Not used for training |
| MistralAI-Training | Crawls to build training datasets | Yes |
So the agent on this page is the live reader. If your concern is training, the rule you want names MistralAI-Training instead, which Mistral says webmasters can disallow.
How to block the MistralAI user fetcher
Mistral says this token governs which sites user requests can reach. That makes robots.txt the right tool, unlike live fetchers from some other companies that say they may skip it.
User-agent: MistralAI-User
Disallow: /To opt out of training only, keep live reads and block the training crawler:
User-agent: MistralAI-Training
Disallow: /Pagereckon Pro offers one-click blocking for bots that crawl hard and give nothing back, such as Bytespider. A live reader that can cite you is not on that list.
What blocking costs
When a Vibe user asks about your topic, Mistral cannot read your page or link it in the answer. Instead, it will draw on other sources. Since the agent does no training, a block buys you nothing on that front.
For most publishers, then, the useful split is the same as with other operators. Allow the live reader and the indexer, and decide about the training crawler on its own.
Verify a MistralAI user request
Mistral publishes the agent's IP addresses at mistral.ai/mistralai-user-ips.json. Take the IP from your log and check that it falls inside one of the listed ranges. The indexer has a list of its own at a similar address.
When the address does not match, the request did not come from Mistral.
MistralAI user reads in Pagereckon
Pagereckon files the agent under live reads, next to ChatGPT-User and Claude-User. Every read stands for a person who asked Mistral's assistant something your page could answer. Visits that arrive from mistral.ai count as AI referral traffic in the same report.
Then compare reads and visits per page. See the AI crawler checker for the live split, or AI referral traffic for the people who clicked through.
MistralAI user agent questions
Does MistralAI-User train Mistral's models?
No. Mistral says it is not used to crawl content for generative AI training. MistralAI-Training does that job.
Does the MistralAI user fetcher obey robots.txt?
Mistral says the token governs which sites these user requests can be made to, so a Disallow is the way to stop it.
Which product sends it?
Mistral's robots page names Vibe, its assistant.
Is tracking it free?
Yes. Live reads are part of the free plugin. Download the free plugin to see them.