Promptwatch Logo

Haiku-SearchBot

The user-triggered fetcher behind Haiku, a French legal AI assistant, which reads web pages to answer a user's question.
Haiku (Clerk SAS)Haiku-SearchBot
AI AssistantUser InitiatedAI (No Training)

What is Haiku-SearchBot?

Haiku-SearchBot is the fetcher Haiku uses for direct user requests. Haiku is an AI assistant for legal professionals, run by the French company Clerk SAS. Its robots documentation says that when a user asks a question, Haiku may run a web search through Google's Vertex AI and then fetch the result pages so it can work their content into the answer and show the user where it came from.

Haiku says Haiku-SearchBot does not explore the web automatically and is not used to collect content for training generative AI models. It also says Haiku does not record information from third-party sites when the request comes from a user, and that its search follows a whitelist policy rather than behaving like a general search engine.

A second identity, HaikuBot, indexes institutional legal data such as court decisions published as open data. It is not the bot that visits ordinary websites for user questions, and Haiku says it is not used for model training either.

The documented user agent is Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; Haiku-SearchBot/1.0; +https://haiku.fr/robots-doc. Haiku follows Cloudflare's Web Bot Auth scheme, and its signing keys are published at https://harvest.haiku.fr/.well-known/http-message-signatures-directory. It supports HTTP/1.1 and HTTP/2, defaulting to HTTP/1.1, and says a server can refuse HTTP/2 by answering with a 421 status.

Relevant for AI searchHaiku

Is Haiku-SearchBot relevant for AI search?

Yes. Haiku-SearchBot fetches pages when users of the Haiku legal AI assistant ask a question, and Haiku shows the sources it used.

Haiku is a French AI assistant for legal work. When a user asks something that needs the web, Haiku runs a search and fetches the result pages with Haiku-SearchBot, then displays the page titles so the user can trace each source. Pages it cannot fetch cannot be used or cited in that answer. Haiku recommends allowing the bot if you want to appear in those results, particularly for legal content.

How to handle Haiku-SearchBot

Keep Haiku-SearchBot allowed if legal professionals are part of your audience. Haiku recommends it for sites that want to appear in its results. To exclude the whole site, or a section, use the token in robots.txt:

User-agent: Haiku-SearchBot
Disallow: /

Haiku tells site owners to exclude sections this way. Its web search also runs through Google, so your Googlebot rules shape which pages Haiku finds in the first place. Verify genuine requests with the Web Bot Auth signature rather than the user-agent string.

Examples

  • A law firm keeps Haiku-SearchBot allowed so its articles on employment law can be used and cited when lawyers ask Haiku about recent rulings.
  • A legal publisher sees single Haiku-SearchBot requests for individual commentary pages and matches them to user questions rather than a crawl.

Frequently asked questions about Haiku-SearchBot

Learn about AI visibility monitoring and how Promptwatch helps your brand succeed in AI search.

Be the brand AI recommends

Monitor your brand's visibility across ChatGPT, Claude, Perplexity, and Gemini. Get actionable insights and create content that gets cited by AI search engines.

Promptwatch Dashboard