What are PerplexityBot and Perplexity-User?
PerplexityBot crawls the web to build the index Perplexity searches, and Perplexity-User fetches a page at the moment a person asks Perplexity a question. On tollbit.com the split runs the opposite way from most assistants: the indexer made 1.9K reads between Sep 15 and Oct 14 2026 and the user-triggered agent made about 0.4K. Perplexity is index-heavy and retrieval-light.
Two agents, two jobs
| PerplexityBot | Perplexity-User | |
|---|---|---|
| Job | Build and refresh Perplexity's search index | Fetch a page live for a user's question |
| User-agent token | PerplexityBot |
Perplexity-User |
| robots.txt | Honored, per Perplexity's documentation | "Generally ignores" robots.txt, per Perplexity's documentation, because a person requested the fetch |
| IP verification | Published JSON list of IP ranges | Published JSON list of IP ranges |
| Reads on tollbit.com, 30 days | 1.9K | about 0.4K |
Perplexity publishes both user-agent strings and both IP lists on its crawler pages at perplexity.ai, which puts it among the better-documented operators. The robots.txt position for Perplexity-User is stated plainly in Perplexity's own docs and matches what Google says about Google-Agent: a user-triggered fetch is treated as a browser request, not a crawl. If you need to control Perplexity-User, do it at the server, not in robots.txt.
What they did on our site
PerplexityBot's 1.9K reads look like an indexer at work: sweeps across the site, revisits to the same pages, no clustering around the questions of the week. Perplexity-User barely appears. About 50 of its roughly 400 fetches were of /careers, which suggests a handful of people asking Perplexity about working at TollBit rather than any sustained retrieval of our content.
Compare the three user-triggered fetchers we see most: ChatGPT-User 10.5K, Claude-User 3.2K, Perplexity-User 0.4K. Perplexity answers more from what it has already indexed and fetches live less often.
Referrals follow the same shape. In the period Perplexity sent 3 human clicks to tollbit.com, against 37 from ChatGPT and 7 from Claude.
The most accurate links of the major assistants
Perplexity has the lowest hallucinated-link rate we have measured: 1.16% of clicks from Perplexity answers land on a 404, against 3.7% for OpenAI in Q2 2025 and 4.8% for Anthropic in the same quarter (our Bits and Bots analysis). Perplexity's design leans on explicit citations and live sources, so the URLs it shows users are more often URLs that exist. For a site owner that means a Perplexity citation converts to a real visit more reliably than a citation from an assistant that guesses.
What index-heavy means for you
If Perplexity reads mostly from its index, what matters is what the index holds. Three things follow:
- Your bot-served sitemap is the lever. PerplexityBot discovers pages the way search indexers do. Expired URLs, missing pages and agent-only pages absent from the sitemap bots read are absent from Perplexity's index too.
- Freshness runs on the indexer's schedule, not the user's. A price or spec change reaches Perplexity when PerplexityBot comes back. That interval varies by platform, from hours to weeks.
- Verification is easy. Perplexity publishes IP ranges, so a request carrying either token from an address outside those ranges is not Perplexity. Check at the edge, before serving.
Allow PerplexityBot as you would any AI search indexer. Allow and measure Perplexity-User as retrieval. TollBit's Bot Index tracks both across 10,000+ sites, and the bot paywall applies per-agent terms to either one.
Frequently asked questions
Can I block Perplexity-User with robots.txt?
Not reliably. Perplexity's documentation says Perplexity-User generally ignores robots.txt because a person asked for the page. A disallow for PerplexityBot does take you out of the index. Enforcement for the user-triggered agent happens at the server, by user agent plus IP.
Why does Perplexity-User appear so rarely in my logs? Because Perplexity composes answers largely from its own index and fetches a page live less often than ChatGPT or Claude do. On our site the ratio was 1.9K index reads to 0.4K live fetches. A site with fast-changing content may see more live fetches.
Does Perplexity pay for content? Perplexity has announced revenue programs for publishers, with terms set between Perplexity and each publisher. Separately, any site can price PerplexityBot fetches itself with an HTTP 402 response through a bot paywall, which does not depend on a program from the operator.