How to get cited by Perplexity: how it picks the sources it shows

How to get cited by Perplexity, in short

You get cited by Perplexity by being reachable, being in its search results, and having the passage that answers the question. The first part you control fully. The other two you can only improve.

Perplexity is an answer engine. It searches the web for each question, reads some of the pages it finds, writes an answer, and puts numbered links to its sources beside the text. A citation is one of those numbered links pointing at your page. It is the closest thing Perplexity has to a ranking, and it can send a visitor.

1. Allow PerplexityBot in your robots.txt and in your firewall.
2. Make sure the page that should be cited is public, loads without a login, and is linked from your other pages.
3. Put a direct answer to the buyer’s question near the top of that page.
4. Back the answer with numbers, named sources and short quotations.
5. Ask Perplexity the question yourself, record which pages it cites, and repeat over several weeks.

The rest of this guide explains each step, what the evidence for it is, and where the evidence runs out.

How Perplexity chooses the sources it cites

Perplexity cites pages from its own search results, and it fills those results with its own crawler. It has not published the rules it uses to rank one page above another.

Perplexity’s crawler documentation names two user agents. A user agent is the name a program gives when it asks your server for a page. Each one does a different job, and each one answers to your robots.txt differently.

Put together, the picture is this. PerplexityBot builds the index of pages Perplexity can find. When someone asks a question, Perplexity searches that index, and Perplexity-User may fetch a page live to read it. The answer then links the pages it drew on. If PerplexityBot is blocked, your page is unlikely to be found in the first place, even though the live fetcher might still reach it.

Perplexity also sells access to its search results through a Search API, which it describes as returning “real-time, ranked web results”. That confirms the order of results is ranked, but the documentation does not say what the ranking rewards. Any article that lists Perplexity’s exact ranking factors is describing a guess.

Step one: let Perplexity reach your pages

Allowing PerplexityBot is the one step Perplexity itself asks for. Its 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”.

1. Open yoursite.com/robots.txt in a browser and read it.
2. Look for a group that starts with “User-agent: PerplexityBot”. If it says “Disallow: /”, Perplexity’s crawler is blocked from the whole site.
3. Look for a group that starts with “User-agent: *”. A “Disallow: /” there blocks every crawler without a group of its own, PerplexityBot included.
4. Ask whoever runs your hosting or content delivery network whether it blocks AI crawlers. A firewall rule can turn PerplexityBot away even when robots.txt allows it.
5. If your firewall needs an allow rule, match both the user agent name and the IP addresses Perplexity publishes, as its documentation suggests.
6. Wait at least 24 hours after any change before you test again.

The firewall step matters more than it used to. On 1 July 2025 Cloudflare announced it was changing its default to block AI crawlers unless they pay creators. If your site sits behind Cloudflare, check its AI crawler setting rather than assuming the robots.txt file is the whole story. Our page on how Perplexity is asked and read covers its crawlers from the assistant’s side.

Step two: write the passage Perplexity can lift

The best public evidence says pages with quotations and statistics get used more by Perplexity, and pages stuffed with keywords do not. That evidence comes from one research paper, with real limits.

The paper is GEO: Generative Engine Optimization, first posted as a preprint in November 2023 and accepted at the KDD 2024 conference. Its authors rewrote source pages in several ways and measured how much of each page an AI answer used. Most of their tests ran on an engine they built themselves. They also ran a smaller test on Perplexity, and those are the figures below.

The authors report that adding quotations did best on their main measure, about 22% above the unchanged page, and that adding statistics improved the second measure by up to 37%. Keyword stuffing did about 10% worse on the main measure. Its score on the second measure went up, which is a reminder that the two measures can disagree.

In practice, a passage Perplexity can lift looks like this. It answers one question in its first sentence. It gives a number and says where the number came from. It quotes a named person or document where that helps. It sits under a heading that says what the section is about. None of this is a trick. It is also what makes a page easy for a person to trust.

Perplexity cites a passage, not a website. Write the paragraph that answers the question, and put it where a reader would look first.

Step three: give it a reason to pick you over the others

When several pages say the same thing, the one with something the others lack has a better chance. This step is our advice, drawn from the research above, not a rule Perplexity has published.

• Own the fact. If you publish the original figure, the price or the specification, you are the natural source for it.
• Date the page. A visible date tells a reader, and a machine, how current the facts are.
• Name the alternatives fairly. A buyer asking for options gets a list, and a page that compares honestly can be cited for the whole comparison.
• Keep one question to one section. A section that wanders across three questions is harder to quote for any one of them.
• Earn mentions on pages Perplexity already cites. Review sites, forums and industry lists that appear in its answers are part of the evidence it reads.

Why a good page still goes uncited

A page can be accurate and well written and still not be cited, because Perplexity never found it or found something it preferred. The causes below are our reading of the documentation and research above, not a list Perplexity has published.

• The crawler was turned away. A robots.txt rule or a firewall setting stopped PerplexityBot, so the page never entered the index.
• The page answers a different question. It covers the topic, but not the question the buyer typed, so another page matches more closely.
• The answer is buried. The useful sentence sits halfway down, after an introduction about the company, where a passage-level reader is less likely to pick it.
• Nothing on it is unique. It repeats what ten other pages say, so any of them would do, and the one Perplexity picks is rarely yours.
• The text is hidden behind a script, a login or a cookie wall. A person sees it after a click, but a crawler reading the raw page may not.

Each cause points to a different fix, which is why checking access comes first. Rewriting a page Perplexity cannot reach changes nothing.

What you cannot control

You cannot control how Perplexity ranks pages, what its live fetcher reads, or whether its citations are right. Plan around all three.

The live fetcher is the first gap. Perplexity’s own page says Perplexity-User “generally ignores robots.txt rules”. In August 2025 Cloudflare reported that it had also seen Perplexity reach sites through a crawler that did not declare itself, posing as Chrome on a Mac, at 3 to 6 million requests a day, against 20 to 25 million a day for the declared Perplexity-User agent. Cloudflare removed Perplexity from its list of verified bots. Perplexity disputed the account in a reply titled “Agents or Bots? Making Sense of AI on the Open Web”.

Accuracy is the second gap. In March 2025 the Tow Center for Digital Journalism gave eight AI search tools excerpts from news articles and asked them to name the source. Perplexity answered 37% of its queries wrongly, against 94% for Grok 3. The researchers also found that Perplexity’s free version correctly identified all ten excerpts they shared from paywalled National Geographic articles, even though, they write, the publisher has disallowed Perplexity’s crawlers. A citation to your page is evidence Perplexity used it. It is not proof the answer describes your page correctly.

How to check whether you are being cited

Ask Perplexity the questions your buyers ask, write down which pages it cites, and repeat the same questions every week. Change shows up as a pattern across weeks, not in one answer.

The pages cited instead are the most useful column. They tell you who Perplexity trusts on that question, and opening them shows what they say that your page does not. Our method page sets out the counting rules we follow, such as never blending branded and non-branded questions.

Agentcites runs this log for you. It asks up to 50 buyer questions to Perplexity, ChatGPT, Gemini and DeepSeek every week, keeps each answer word for word, and records every page each answer linked. It also checks your robots.txt for five AI crawlers, PerplexityBot among them, and quotes the line that blocks one. It asks through each assistant’s API, so an answer can differ from what a signed-in person sees. The plan is $149 a month per website, on the pricing page.