How to recognize it

User agent, as it reaches our logs (identical to Perplexity’s docs):

Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; PerplexityBot/1.0; +https://perplexity.ai/perplexitybot)

Match on the token PerplexityBot, then check the IP: 88% of requests with this user agent on our sites were fake.

  • robots.txt token: PerplexityBot
  • IP list (JSON): perplexitybot.json. Perplexity’s docs link it under perplexity.com, which redirects here.
  • Reverse DNS: none published by Perplexity.

What it does

PerplexityBot finds and links websites for Perplexity’s search results. Perplexity says it is not used to crawl content for training AI foundation models. Its docs tell site owners to manage it through robots.txt and recommend allowing it to appear in search results. They do not mention Crawl-delay.

What you gain by allowing it

  • Appearing, with a link, in Perplexity’s answers, and the clicks they send.

What you lose

By allowing it

  • Little load: 101 verified requests in 8 days on our two sites.
  • Perplexity’s answers can summarise your page well enough that the reader does not click through.

By blocking it

  • Your pages drop out of Perplexity’s search results.
  • It may not stop Perplexity entirely. Perplexity-User, which fetches pages for users, generally ignores robots.txt. In 2025, Cloudflare also reported that Perplexity fell back to an undeclared browser user agent when blocked.

How to verify it

Match the IP against perplexitybot.json; the one-liner on the GPTBot card works with this URL. Perplexity itself recommends checking the user agent together with the IP list in a firewall.

Do not rely on Cloudflare’s verified-bot flag. Cloudflare removed Perplexity from its verified bots in August 2025, so a rule that lets only cf.client.bot through treats the real PerplexityBot like a fake.

Gotchas and history

  • Cloudflare’s 2025 report: on 4 August 2025 Cloudflare reported that when Perplexity’s declared crawler was blocked, Perplexity switched to a Chrome-on-Mac user agent and addresses outside its lists, sometimes ignoring robots.txt. Cloudflare de-listed it as a verified bot and added rules against the undeclared crawling.
  • Propagation: Perplexity says robots.txt changes can take up to 24 hours to apply.
  • Test bed versus our sites: on our test bed the real PerplexityBot fetched only robots.txt and the homepage; see One IP claimed to be Google, OpenAI, Anthropic, Apple and Meta. On our other sites it read articles too (see below).

What we observed

Data: Cloudflare Logpush for two of our sites, 19-26 October 2026 (8 days).

  • 828 requests carried a PerplexityBot user agent:
    • 101 (12%) came from Perplexity’s list, all from Amazon’s network (AS14618);
    • 718 of the 727 fakes came from the Google Cloud scanner.
  • Cloudflare did not flag any of the 101 listed requests as a verified bot.
  • 24 of the 101 were for /robots.txt, 3 for a homepage and 74 for other pages, on both sites, on all 8 days.
  • 6 got a 4xx from the origin server, 5 of them on the WordPress site.

About Perplexity

  • PerplexityBot finds and links websites for Perplexity’s search results. Perplexity says it is not used to collect content for training foundation models.
  • Perplexity-User opens a page when a user’s question needs it, and may link to that page in the answer. Perplexity says it generally ignores robots.txt, because a user asked for the fetch.

Perplexity tells site owners to manage both through robots.txt, and recommends checking the user agent together with the IP lists in a firewall. It says robots.txt changes can take up to 24 hours to apply.

The docs link the lists under perplexity.com, which redirects to perplexity.ai.

In August 2025 Cloudflare reported that Perplexity switched to an undeclared browser user agent and unlisted IPs when its declared crawler was blocked. Cloudflare removed Perplexity from its verified bots. On our sites, Cloudflare still did not flag a single PerplexityBot request as verified in October 2026, including those from Perplexity’s own list.

Sources