Quick Answer: PerplexityBot is Perplexity's search crawler that indexes and links websites in Perplexity results, and it is not used to train AI models. Perplexity-User is the separate agent that opens a page when a person asks a question. Blocking PerplexityBot removes you from Perplexity's search results; changes take up to 24 hours to apply.
Perplexity documents exactly two robots, each with its own robots.txt token, a full user-agent string and a published IP range, in its crawler guide. Unlike OpenAI and Anthropic, Perplexity states that neither bot collects content for training foundation models, which changes the blocking decision for most sites.
What PerplexityBot does and does not do
PerplexityBot crawls and indexes pages so they can be surfaced and linked in Perplexity's search results. Perplexity states it "is not used to crawl content for AI foundation models." Its published full string is Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; PerplexityBot/1.0; +https://perplexity.ai/perplexitybot).
Perplexity's recommendation is direct: allow PerplexityBot in robots.txt and permit requests from the published IP ranges if you want to appear in results. It also notes that robots.txt changes can take up to 24 hours to be reflected, so a test run five minutes after an edit proves nothing.
At a glance
| Agent | Job | Used for training | Effect of blocking |
|---|---|---|---|
| PerplexityBot | Indexes and links sites in Perplexity search results | No | Site stops appearing in Perplexity results |
| Perplexity-User | Visits a page to answer a user's question and link it | No | Perplexity cannot open your page for that user |
What it is not
PerplexityBot is not a training crawler, so the "keep us out of the model" argument that applies to GPTBot and ClaudeBot does not apply here. Blocking it buys no training protection and costs every Perplexity citation your site could have earned.
It is also not tied to Perplexity-User. Perplexity states that each robots.txt setting "works independently", so allowing one agent and blocking the other is a valid configuration, and a rule written for one token is ignored by the other.
PerplexityBot vs Perplexity-User
PerplexityBot builds the index in advance; Perplexity-User opens a page live because a person's question needs it. Perplexity describes the second as supporting "user actions within Perplexity" and says it "might visit a web page to help provide an accurate answer and include a link to the page in its response."
The distinction decides where you look when citations drop. If Perplexity-User requests stop arriving, users' questions are no longer reaching your pages, which is usually an index or firewall problem; if they arrive and no citation follows, the page is being read and not quoted.
The user-agent strings
Perplexity-User carries the token Perplexity-User/1.0 inside a browser-style string. The published form is Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; Perplexity-User/1.0; +https://perplexity.ai/perplexity-user). Match on the tokens PerplexityBot and Perplexity-User in logs, and verify the source IP against Perplexity's published ranges before trusting either name.
Why Perplexity cites more pages than ChatGPT
Perplexity searches on nearly every question and cites several sources per answer, which is why sites often see more Perplexity-User traffic than ChatGPT-User traffic. Ranking in Perplexity is therefore a citation-first game, and access is the first gate.
Should you block PerplexityBot
No, unless you have a specific reason to keep your content out of Perplexity's search results altogether. Because neither Perplexity bot is used for training, the usual privacy or licensing case for blocking does not apply, and the cost is measurable in lost citations.
The case for allowing
Perplexity is one of the four engines most AI visibility programmes track, and it leans heavily on freshly indexed pages. A site that is invisible to PerplexityBot is invisible on a surface that sends real referral sessions, which AI traffic reports in GA4 can show you week by week.
The case for rate limiting
Rate limit rather than block when PerplexityBot hits a large site too hard. A Crawl-delay directive or a WAF rate limit on the published IP ranges keeps the index alive while protecting the origin.
The same logic applies to any crawler that earns you citations. A block is permanent until someone remembers to lift it; a rate limit is a dial you can turn, and answer engine optimization depends on the dial staying open.
When a block is legitimate
Paywalled or members-only content that should never be summarised is the clear case. Disallow the specific paths for both agents rather than the whole site, so public pages keep their visibility.
Staging and preview hosts are the other case. A preview subdomain that is reachable without a login will be indexed like any other host, so give it its own robots.txt with both agents disallowed, and keep the production file open.
How to configure a WAF so Perplexity is not blocked by accident
Whitelist Perplexity's published IP ranges and user agents in your firewall, because managed bot-protection rules on Cloudflare and AWS routinely challenge both agents. Perplexity's doc includes step-by-step rules for both providers and recommends verifying against the IP source list rather than the user agent alone.
Cloudflare
Under Security, then WAF, create a custom rule that skips managed challenges when the user agent contains PerplexityBot or Perplexity-User and the source IP is in Perplexity's published list. Cloudflare's own "AI bots" managed toggle may otherwise block both, so check its scope.
AWS WAF
Add an allow rule ahead of your bot-control rule group that matches the two user-agent tokens and the published IP set. Order matters: an allow rule placed after a block rule never fires.
Testing the change
Wait 24 hours, then ask Perplexity a question your page answers and look for the page among the cited sources. The robots.txt reference for AI bots has the equivalent tokens for the other engines if you want to run the same test everywhere.
How to read PerplexityBot activity in your logs
Filter 30 days of access logs for each token separately and chart requests and status codes by day. Healthy sites see steady PerplexityBot crawling across the whole site and bursts of Perplexity-User requests on the specific pages that answer live questions.
Three patterns and what they mean
- PerplexityBot 200s, no Perplexity-User requests. You are indexed but never chosen; the answer block on the page is not competitive, so start with how the page is structured.
- Perplexity-User requests, no citations. The page is opened and discarded; the claim it makes is not specific or attributable enough to quote.
- 403s on either agent. The firewall is refusing them, and no amount of content work will help until it stops.
Pairing the log with a citation check
Perplexity-User requests are the closest thing to a citation log that a server can produce on its own. Keep a monthly list of the pages Perplexity-User fetched and mark which ones Perplexity actually cited. The gap between the two lists is the most precise content brief you will get, and measuring AI visibility per page is what keeps it current.