How to Rank on Perplexity: What Its Own Documentation Supports

Key takeaways

The short version of this page, in the order the sections below develop it. Each point is stated here in full so it can be read without the rest of the page, and the section that argues it gives the measurement and the source behind it.

• Perplexity recommends allowing PerplexityBot in robots.txt and permitting its published IP ranges.
• Perplexity says PerplexityBot will not index the text of a site that disallows it in robots.txt, but that it may still index the domain, the headline and a brief factual summary of a blocked page.
• Perplexity's research article on its Search API says documents are decomposed into self-contained spans that are retrieved and ranked individually. The documents we read do not say how spans are weighted.
• In our S&P 100 study (2026-09-25, 96 answers per assistant), Perplexity's sonar model, the one assistant that searched the live web, named 34 companies; Gemini named 46 and Claude 39. ChatGPT named 36, from only 54 of 96 answers.

How do you get your company cited in Perplexity answers?

You make your pages reachable by Perplexity's crawler, which its documentation covers, and write passages that answer one question by themselves, which is our inference; nothing we read guarantees a citation. Perplexity's help center article How does Perplexity work? (archived copy of 2026-09-24) says "Each answer includes numbered citations linking to the original sources".

1. Allow PerplexityBot in robots.txt. Perplexity's crawler documentation says: "To ensure your site appears in search results, we recommend allowing PerplexityBot in your site's robots.txt file and permitting requests from our published IP ranges listed below."
2. Let it through the firewall. The same page warns that with a web application firewall "you may need to explicitly whitelist Perplexity's bots". It advises combining user-agent matching with IP address verification, and gives the procedure for Cloudflare WAF and AWS WAF.
3. Leave Perplexity-User alone. The same page says that when users ask a question, this second agent "might visit a web page to help provide an accurate answer and include a link to the page in its response." Perplexity's recommendation to allow names PerplexityBot only; leaving Perplexity-User unblocked is our advice. The firewall instructions cover both agents.
4. Write passages that stand alone (our reading, not Perplexity's advice). In an engineering article about its Search API infrastructure, Perplexity's research team says its parser lets it "decompose documents into self-contained spans, each of which can be individually retrieved and ranked at query time." It gives site owners no instructions. Our inference: put the question in the heading and the full answer directly under it.
5. Keep the page current, then wait (our reading, not Perplexity's advice). The same Search API article says prefiltering stages "remove clearly non-responsive or stale content from the candidate set". It does not define stale; keeping a page current is our inference. For robots.txt settings, the crawler documentation says "it may take up to 24 hours for our systems to reflect changes." The documents give no interval for re-reading an edited page.
6. Check the logs, then ask. After a firewall change, the crawler documentation says to "Monitor your logs to ensure the rules are working as expected". Then ask Perplexity the questions your buyers ask and record which URLs it cites. Our method for checking whether an assistant mentions your brand applies unchanged.

Steps 1, 2, 3 and 6 can be checked in your own robots.txt, firewall rules and logs. Steps 4 and 5 cannot: they are UPAI's reading of an article about API infrastructure, not Perplexity's advice to site owners.

What does Perplexity say about how it picks sources?

Perplexity describes a search index of its own, a parser that cuts pages into spans, and several ranking stages that end in reranker models; the documents we read do not give the signals or their weights. The fullest account is a Perplexity Research article dated Sep 25, 2025, "Architecting and evaluating an AI-first Search API", which describes the infrastructure behind an API sold to developers. We read it in an archived copy dated 2026-09-30, because perplexity.ai refused our automated requests.

Two limits apply to this table. The article describes API infrastructure: a figure caption says Perplexity is "powered by" that search infrastructure, but the text is an engineering description published to launch an API, not guidance for site owners. And "tracks" a URL is Perplexity's word; it is not the same claim as holding a current copy of every page.

What happens if you block PerplexityBot?

Perplexity says it will not index the text of your pages, but that it may still index the domain, the headline and a brief factual summary. The wording comes from its help center article How does Perplexity follow robots.txt?, which we read in an archived copy dated 2026-04-01. On 2026-10-05 we compared it with the current article, reached indirectly, and the sentences quoted here were unchanged.

The article does not say how a blocked page's headline and summary are used in answers.

The second row needs a caveat. The help center article says a feature that let users have Perplexity summarize a URL blocked by robots.txt "has been disabled to prevent misuse", while the crawler documentation says Perplexity-User generally ignores robots.txt. We cannot tell from the two which describes current behavior.

None of the documents we read says how Perplexity treats a page behind a login or a paywall.

Does being crawled mean Perplexity will name your company?

Being crawled does not mean Perplexity will name your company: access makes you eligible, and in the one measurement we have, the assistant that searched the web named fewer companies than those that did not. In our S&P 100 study, run on 2026-09-25, four assistants answered the same unbranded buying questions. Perplexity's sonar model, called through the API, was the only one that searched the live web. In the eight sectors we could analyse, it named 34 distinct S&P 100 companies.

Gemini named 46 and Claude named 39, both from what they had learned, with no search. All three returned text in 96 of 96 answers, so that comparison is like for like. ChatGPT named 36, but returned text in only 54 of 96 answers; fewer answers means fewer chances to name a company, and we cannot say by how much that lowered its count.

• It shows that live retrieval did not produce a wider set of names in this test.
• It does not show why. We did not measure whether sonar's answers were shorter or whether it kept to the companies its sources named.
• It counts company names, not cited URLs. Being named and having your page cited are different observations.
• It covers S&P 100 companies, on the sonar model. It is not a figure for mid-size firms, nor for the Perplexity consumer app. The other three were also API models.

What is still unknown about ranking on Perplexity?

The documents we read do not say what decides who gets cited, so we could not trace any weighted list of "Perplexity ranking factors" to them. We read five: the crawler documentation, two help center articles, the research article and the Publishers' Program announcement. The list below is what those five do not say, not a statement about how Perplexity works:

• Whether PerplexityBot runs JavaScript. None of the five documents says. Serving the answer in the HTML is the cautious choice, not a documented requirement.
• Whether links count. None of the five documents names links or brand mentions as a ranking signal.
• How the answer model picks among retrieved spans. The research article describes the pipeline up to the ranked result set and no further.
• Whether a commercial relationship changes citation. The Publishers' Program announcement of July 30, 2024 (archived copy of 2026-07-18) says that when Perplexity earns revenue from an interaction where a publisher's content is referenced, "that publisher will also earn a share." It does not say partners are cited more often, and we have not checked whether those terms still apply.
• How stable a citation is. We have not measured how often the cited sources change when the same question is asked again.

Tracking vendors can count citations from outside, which is not the same as knowing the ranking; our comparison of tools that track brand visibility in AI search covers what they observe. We sell AI-visibility work ourselves, so apply the same test to us.

Questions people ask next

These are the follow-ups we are asked most often once the answer above has landed. The same rule applies to each: where a figure appears, its source and year sit in the same sentence, and where we do not know something we say so rather than estimate it.

Does Perplexity use Google or Bing to find sources?

Perplexity describes a search index of its own. Its research team wrote in September 2025 that the very first version of the answer engine incorporated some third-party search APIs, and that the company then concluded it needed its own dedicated search infrastructure. Its help center adds that it partners with unnamed third-party crawlers to help build the index. None of the documents we read says it queries Google or Bing today.

Will allowing PerplexityBot let Perplexity train AI models on my content?

Not according to Perplexity. Its crawler documentation says PerplexityBot is not used to crawl content for AI foundation models, and its help center says Perplexity does not build foundation models, so content will not be used for AI model pre-training. Both are the company's statements; we cannot audit them.

How long does it take Perplexity to pick up a change on my site?

For robots.txt changes, Perplexity's crawler documentation says it may take up to 24 hours for its systems to reflect them. For page content, the documents we read give no interval. The research article says a machine-learning model decides whether a URL needs indexing, calibrated to its importance and likely update frequency.

Can you pay to be cited in Perplexity answers?

We found no Perplexity document that offers paid placement among an answer's citations. The Publishers' Program announcement of July 30, 2024 describes revenue sharing with partner publishers and a plan for advertising in which brands pay to ask specific related follow-up questions. It does not say partners are cited more often. We read that 2024 text in an archived copy and have not verified what Perplexity sells today.

Related

Other questions in this library that cover the ground next to this one. Each is answered the same way — the direct answer first, then the measurement behind it — and each carries its own sources, so you can check any of them on its own page without taking this one on trust.

• GEO vs SEO: Which Work Carries Over and Which Does Not