How Perplexity cites sources (from Perplexity's own docs)
Short answer: Perplexity searches the web for each question, writes an answer, and attaches numbered citations that link to the original sources. Its crawler PerplexityBot finds and links sites for results (not for model training). A second agent, Perplexity-User, fetches pages when a user asks and generally ignores robots.txt. Some cited domains carry a Government, Academic or Trusted label, which rates the whole site and, by Perplexity's account, isn't affected by payments. Perplexity doesn't publish a formula for which pages get cited, and this guide doesn't guess one.
Sources checked on 15 November 2026. Links are at the end.
What Perplexity's help center says
- How answers are made: Perplexity searches the internet in real time, gathers information from sources such as articles, websites and journals, and summarises it.
- Citations: each answer includes numbered citations linking to the original sources, so readers can verify the information.
- Source labels: some citations show a shield icon with a Government, Academic or Trusted label. Perplexity rates the whole website, not individual pages, using objective questions such as whether the site corrects its mistakes, says who wrote each piece, and keeps news separate from advertising and opinion. Perplexity says partnerships, payments and other business arrangements don't affect labels, that most domains have no label, and that having no label is not a negative judgment.
What Perplexity's crawler docs say
| Agent | Purpose (per Perplexity) | robots.txt |
|---|---|---|
PerplexityBot |
Surfaces and links websites in Perplexity search results; not used to crawl content for training foundation models | Respected. Allow the user agent and its published IP ranges; changes can take up to 24 hours |
Perplexity-User |
Visits pages to answer a specific user request | Generally ignores robots.txt, because a user requested the fetch |
Perplexity also notes that some web-application-firewall modes (it names Cloudflare Bot Fight Mode) can't be bypassed with custom allow rules.
What the developer docs show about citations
Perplexity's API cookbook for streaming citations describes the mechanics for developers:
- The model inserts numbered markers such as
[1]and[2]into the text. - The source URLs arrive separately as search results, each with an
id,title,url,snippetanddate. Theidmatches the number in the text. - A response can run several searches that share one citation numbering, so all batches of results must be collected.
- The docs warn developers never to ask the model to generate source URLs, because model-generated URLs can be hallucinated, and to validate URLs before displaying them.
In other words, a citation is a pointer to a search result Perplexity actually retrieved. The sentence before it is still generated text, which is why Perplexity's help pages tell readers to check the source.
What we observed ourselves
On 15 November 2026 at 20:59 KST we tried to load a Perplexity answer page in an automated headless browser to record its citations. Perplexity showed a security verification page instead, so we have no first-hand observation of an answer to report from that session. Everything above comes from Perplexity's own documentation. To observe it yourself, ask a question in a normal browser, hover over each numbered citation, and note the domain, any label, and whether the cited page really says what the sentence claims.
What this means for your site
- Don't block PerplexityBot in robots.txt, and check that your firewall's bot protection isn't blocking it.
- Make the claim easy to find in text on the page you want cited. Citations attach to specific statements.
- Put a date and an author on pages where it fits. Search results in the API carry a date, and the source-label review asks whether a site says who wrote each piece.
- Correct mistakes visibly. It's one of the objective questions in Perplexity's source review, and it's good practice anyway.
- Measure by asking. Use the AI search visibility checklist to record citations month by month.
For the cross-platform picture, see how to get cited by ChatGPT, Perplexity and Google AI Overviews. A GEO audit ($149, 48 hours, pricing) includes Perplexity in its buyer-question checks.
Sources
- Perplexity Help Center, "How does Perplexity work?": https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work (read 2026-11-15)
- Perplexity Help Center, "Understanding source labels": https://www.perplexity.ai/help-center/en/articles/20260806-understanding-source-labels (read 2026-11-15)
- Perplexity, "Perplexity Crawlers": https://docs.perplexity.ai/guides/bots (checked 2026-11-15)
- Perplexity Docs, "Streaming Citation Parsing": https://docs.perplexity.ai/docs/cookbook/articles/streaming-citations/README (read 2026-11-15)
Frequently asked questions
How does Perplexity cite sources?
Each answer includes numbered citations linking to the original sources. In the API, numbered markers map to search results that carry an id, title, URL, snippet and date.
Is PerplexityBot used to train AI models?
Perplexity says no: PerplexityBot surfaces and links websites in its results and is not used to crawl content for training foundation models.
Does Perplexity-User follow robots.txt?
Perplexity says it generally ignores robots.txt, because the fetch is requested by a user.
What do Government, Academic and Trusted labels mean?
They rate a whole domain after Perplexity's source review. Perplexity says payments don't affect labels and that no label is not a negative judgment.
Can I pay Perplexity to be cited?
Perplexity's documentation describes no paid citation, and says business arrangements don't affect source labels.
Start: email yjs121256@gmail.com with your site (or product link), what you need and your deadline. You get a fixed written quote, usually within one Korean business day.