How Perplexity chooses sources: what Perplexity documents, and what it does not

The short answer. To be eligible as a Perplexity source, a page must be reachable by PerplexityBot: allowed in robots.txt and let through your firewall, where Perplexity recommends checking both the user agent and its published IP ranges. Perplexity says its answers carry “numbered citations linking to the original sources”, and it now marks some sites with a source label after a review of how they publish. It does not document how it chooses or ranks sources.

Everything below comes from Perplexity’s crawler documentation and two help-centre articles, How does Perplexity work? and Understanding source labels, as read on 15 October 2026.

Which Perplexity agents visit your site?

Two, and only one of them reads robots.txt.

Agent	What Perplexity says it does	robots.txt
PerplexityBot	“Designed to surface and link websites in search results on Perplexity”; “not used to crawl content for AI foundation models”	Honoured; Perplexity recommends allowing it to appear in search results
Perplexity-User	Supports user actions: when someone asks a question it may visit a page and link to it in the answer; not used for crawling or training	Because a person triggered the fetch, it “generally ignores robots.txt rules”

Two consequences follow. Blocking PerplexityBot is a search decision, not a training one: Perplexity lists no training crawler on that page. And a page can still be fetched for a single answer through Perplexity-User even when PerplexityBot is blocked, but it will not be in the index for other questions.

What does Perplexity ask site owners to do?

Three things, all about access.

1. Allow PerplexityBot in robots.txt. Perplexity recommends it so the site can appear in search results, and says robots.txt changes can take “up to 24 hours” to take effect.
2. Let it through your firewall, and verify it. If you run a web application firewall, Perplexity says you may need to allow its bots explicitly, and recommends matching the user agent and checking the request comes from its published IP ranges (PerplexityBot, Perplexity-User). A user-agent match alone can be forged.
3. Keep the IP lists current. The ranges are “updated regularly”, so Perplexity advises fetching them on a schedule and automating firewall updates rather than pasting them in once.

How does Perplexity build an answer?

At the level of detail Perplexity publishes: it searches “the internet in real-time”, gathers information from sources such as “articles, websites, and journals”, and returns an answer with numbered citations. Its Deep Research mode “performs dozens of searches automatically, reads hundreds of sources”. The help page does not explain how sources are selected or ordered.

1. 1

   Crawled

   PerplexityBot fetches the page for the search index.

   Stopped by: a Disallow, or a firewall that blocks its IP ranges

2. 2

   Retrieved

   Found by a real-time search for the question.

3. 3

   Cited

   Shown as a numbered citation linking to the page.

   Not documented: how sources are chosen

4. 4

   Labelled (some sites)

   A shield icon: Government, Academic or Trusted.

   Whole-site review; payments do not affect it

Separately: Perplexity-User

Fetches a page when a person’s question needs it, and generally ignores robots.txt.

What are Perplexity’s source labels?

A shield icon on some citations that tells readers “what kind of source you are reading”. The help article describes three labels: Government (official government sites), Academic (scientific sites) and Trusted (sites that appear often enough in results and publish within their area of expertise). Labels apply to “the whole website a source comes from, not each individual page”, and are set by a review that asks questions including:

• “Does the site correct its mistakes?”
• “Does it say who wrote each piece?”
• “Does it keep news separate from advertising and opinion?”

Perplexity says partnerships, payments and other business arrangements do not affect a label, that a label “is not an endorsement”, and that having no label is not a negative judgment. The article does not say whether labels affect which sources are used or how they rank, and it describes no way to apply for one.

The review questions are still useful as a checklist. They are the same signals readers and other systems use to judge a source: a corrections policy, named authors, and a clear line between content and promotion. Our editorial policy is our own answer to them.

What does Perplexity not document?

• Selection and ranking. Neither help article explains how sources are chosen for an answer.
• A site-level report. Perplexity offers site owners no equivalent of Search Console’s Generative AI performance report or Bing’s AI Performance report.
• Special markup or files. Nothing in the documentation asks for structured data, llms.txt or a dedicated sitemap.

Measure it from the outside instead: Perplexity is one of the three engines in our Lab prompt set, and the sampling method works the same way for it as for the others. On the visit side, analytics can attribute sessions referred by Perplexity. In our case studies the volume is still small: on woodcharcoal.in, for example, GA4 recorded one session from Perplexity against 51 from ChatGPT over the same window.

A checklist

1. robots.txt allows PerplexityBot (check the live file).
2. Your firewall or bot protection lets through requests that match both the PerplexityBot user agent and its published IP ranges, and the ranges are refreshed automatically.
3. You have decided what to do about Perplexity-User, knowing it generally ignores robots.txt; blocking it is a server-side decision.
4. Pages show who wrote them, the site says how it corrects mistakes, and promotion is clearly separated from content.
5. Perplexity is in your sampled prompt set, scored for mentions, citations and accuracy.
6. Calendar a quarterly re-read of the crawler page and the source-label article.

Our AI crawler access tool fetches a URL as PerplexityBot and Perplexity-User and reports what each receives.

What we don’t claim

• We do not claim to know how Perplexity selects or ranks sources. It does not document it in the pages linked here.
• We do not claim a source label can be earned on request or that it improves citations; Perplexity describes neither.
• We do not claim the details are permanent. Vendor documentation changes without notice; re-read the sources before relying on them.