MistralAI-User

• Answering questions: when someone asks Vibe a question, the agent may fetch a relevant page to help answer.
• Source links: Vibe may link to your page in its answer, which can send readers to your site.
• User-triggered retrieval: The agent handles requests users start. Mistral doesn’t describe it as an automatic discovery crawler.
• Separate site-level choices: You can set a robots.txt rule for MistralAI-User independently of MistralAI-Index and MistralAI-Training.

A visit doesn’t mean the page will be cited, that a reader will click through, or that the answer will use its content.

Mistral’s documented purpose explains why requests happen, but request patterns alone can’t show whether a client is genuine. A user-agent string can be copied.

• Spoofing: An unrelated scraper can send the MistralAI-User string.
• Allowlists: Trusting the user-agent without validating the source can let unverified traffic reach resources meant for a legitimate agent.
• Request volume: Many or repeated requests across unrelated pages deserve a look. Volume alone doesn’t prove spoofing, because Mistral publishes no request-rate threshold for this agent.
• Mixed-up agents: The wrong token or IP list leads to the wrong classification. MistralAI-User and MistralAI-Index have separate IP lists.
• Stale IP data: An old copy of the list can misclassify legitimate requests or let unverified ones through.

Requests to login pages, APIs, or other restricted paths deserve review too, but they don’t prove impersonation.

A spoofed user-agent doesn’t give anyone access to Mistral’s systems or a user’s conversation. The impact on your site comes from the requests and how your infrastructure handles them, such as server load, noisy logs, and misapplied access rules.

If MistralAI-User shows up in your logs, the request claims to come from a user-initiated action in Vibe. The user-agent alone can’t show that a real user asked about your site or that Mistral sent the request.

• User-triggered activity: Mistral documents this agent as responding to user actions in Vibe, not crawling the web automatically.
• Page retrieval: Vibe may visit a page to help answer a question. Mistral doesn’t say each request targets only one page or that the agent never makes several.
• Not the training crawler: Mistral says MistralAI-User isn’t used to crawl content for AI training. MistralAI-Training has its own purpose.
• Possible referral traffic: Vibe may link to a source in an answer. A request doesn’t guarantee a citation or a click.
• No fixed schedule: Mistral doesn’t publish a schedule or an expected request rate.

1. Check the user-agent. Mistral documents this string: Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; MistralAI-User/1.0; +https://docs.mistral.ai/robots). Use the MistralAI-User token to classify requests in your logs and policies. It’s not proof of origin, because user-agent strings can be copied.
2. Check the source IP. Mistral links to its MistralAI-User IP list from its crawler documentation. Compare the request’s IP with the current list. A match adds a verification signal. A non-match means the list alone doesn’t verify the request. Don’t rely on the user-agent as your only check.
3. Keep the IP list current. If you depend on the list, fetch it on a schedule and compare it with your copy. If you automate this, monitor for failed or malformed responses and unexpected changes. Define a fallback policy instead of trusting stale data or treating every failed update as malicious.
4. Use the right list. MistralAI-Index has its own IP list, so don’t swap one for the other. Mistral lists no IP file for MistralAI-Training. Check the official page for updates.
5. Look at behavior. Review request volume, requested paths, response patterns, and your access policies. Unexpected activity can justify a closer look, but neither a high request rate nor a normal-looking request confirms identity. Assess requests to restricted paths against your security policies, whatever user-agent they claim.
6. Check your robots.txt. If requests continue after you disallow MistralAI-User, review your deployed file, rule syntax and placement, caching, and the exact user-agent match. Also check where the requests come from. Continued requests don’t show whether a genuine crawler ignores your rules or another client uses the same string.

It depends on your content policies, security needs, and whether you want Vibe to retrieve your pages when people ask questions.

Blocking the agent may stop compliant user-triggered requests from fetching your content. Mistral doesn’t say how that affects an answer, which depends on the question and what the assistant can access otherwise.

Blocking MistralAI-User doesn’t block other search engines. Mistral documents its three agents separately, so you can set a different preference for each token.

Consider restricting or blocking the agent when:

• Requests with this identity fail your verification checks
• Your content is subscription-based, licensed, or restricted for automated access
• Paths such as account or checkout areas shouldn’t be open to crawlers
• Request volume affects performance or infrastructure costs
• Your policy doesn’t allow this kind of user-triggered retrieval

If you want Vibe to retrieve and possibly cite your pages, allowing MistralAI-User may help. It doesn’t guarantee retrieval, citations, or visibility to users. You can also set a separate preference for MistralAI-Training to tell user-triggered retrieval apart from training data collection.

Use the dedicated robots.txt token for MistralAI-User.

Block the whole site:

User-agent: MistralAI-User
Disallow: /

Block selected paths:

User-agent: MistralAI-User
Disallow: /account/
Disallow: /checkout/

These rules state your preferences. As RFC 9309 explains, robots.txt isn’t access authorization, and a client that chooses not to honor it can still request a URL.

Other controls:

• Keep rules separate: Write independent rules for MistralAI-User, MistralAI-Index, and MistralAI-Training. Check wildcard and overlapping groups so the full file matches your policy.
• Verify before allowlisting: Combine the token with a check against Mistral’s current IP list and other signals. A user-agent-only allowlist is easy to bypass.
• Protect restricted resources: Use authentication and authorization for private content. Robots.txt and an IP allowlist aren’t a complete security boundary.
• Set rate limits with care: Rate limiting can control excess traffic. Base thresholds on observed behavior, and don’t assume a low limit will never affect legitimate requests.
• Monitor after changes: Review server, CDN, and bot management data to see whether traffic changes, and look into requests that continue.

MistralAI-User fetches pages on demand when a person asks Mistral’s assistant a question, and it can cite your page in the answer. Mistral says it isn’t used for automatic crawling or AI training. It documents its user-agent, a robots.txt token, and published IP addresses.

Our advice: this is on-demand, user-driven traffic, so treat it as a visibility decision rather than a threat. Allow it where you want to be cited, and limit or block it on pricing, account, and checkout paths. Verify the user-agent against Mistral’s published IPs, and watch its behavior, since a legitimate identifier doesn’t guarantee a legitimate request.

This is where we can help. With DataDome Bot Protect, you can see and control Mistral user-triggered traffic with intent-based detection, and decide which sections it can reach. Our Agent Trust capability scores each agent in real time and keeps checking its behavior across the session.

Whatever you decide, recheck Mistral’s documentation as its products evolve, and protect sensitive paths with authentication and authorization.