PerplexityBot vs. Perplexity-User: which should you block?

Perplexity documents two crawler tokens with a clear, stated separation: one for indexing, one for answering a specific question — and, importantly, only one of them reliably respects robots.txt.

The two tokens

• PerplexityBot — search. Perplexity states this crawler is “designed to surface and link websites in search results on Perplexity,” is explicitly not used to train AI foundation models, and respects robots.txt.
• Perplexity-User — user-triggered. Fetches a page when a person asks Perplexity a question that requires visiting it directly, and Perplexity’s own documentation states it generally ignores robots.txt because a specific person, not a bulk crawl, requested the fetch.

The decision

• Disallowing PerplexityBot removes your site from being indexed and cited in Perplexity’s search answers — this is the closest equivalent to blocking a conventional search engine, and this rule is actually honoured.
• Disallowing Perplexity-User is not a reliable way to prevent Perplexity from fetching your page on a user’s direct request — Perplexity’s documentation states this fetcher generally ignores robots.txt precisely because a real person asked the question. If you need to prevent that specific fetch, a robots.txt rule alone is not sufficient; see /limitations for what a robots.txt rule can and cannot guarantee.
• Because Perplexity states neither token is used for foundation model training, there is no separate “training opt-out” decision to make for this operator, unlike OpenAI, Anthropic, Google, or Meta.

Common mistake CrawlPact flags

Site owners who want to “opt out of AI training” sometimes disallow PerplexityBot expecting a training effect, then are surprised their site loses Perplexity search visibility with no training-related benefit, since Perplexity’s own documentation states this crawler isn’t used for that purpose. Confirm what a token actually does — via the crawler directory — before writing a rule intended to achieve a specific outcome, or check your own robots.txt directly with the AI crawler checker.