Perplexity SEO: how Perplexity picks sources and what you can do

What does “Perplexity SEO” mean?

Perplexity SEO has two meanings: getting your pages used as sources inside Perplexity’s answers, and using Perplexity as a research tool for SEO work. This guide is about the first one; Perplexity’s own page “Improve SEO strategy” covers the second.

Searches for “how to rank on Perplexity” assume a rulebook, and there is very little of one. Perplexity explains its bots in detail and its choice of sources hardly at all, so any advice here, ours included, is partly inference. This page keeps apart what Perplexity says on its own pages, read on November 4, 2026, and what nobody outside the company knows. None of it can make Perplexity cite you. The aim is to remove what stops it, and to see clearly what it does.

How does Perplexity work, by its own account?

Its help page “How does Perplexity work?” (last modified October 26, 2026) gives three steps. It interprets your question, searches the web, “gathering information from authoritative sources like articles, websites, and journals”, and compiles an answer. Then comes the part that matters to you: “Each answer includes numbered citations linking to the original sources.”

So you can see the pages behind an answer nearly every time you ask. Our guide to AI citations covers how the four assistants show their sources. How many pages you see here depends on how the question was asked.

Under the answers sits its own search index, which a Perplexity research post dated September 25, 2025, “Architecting and evaluating an AI-first Search API”, says “tracks over 200 billion unique URLs”. The system searches by keywords and by meaning, removes “clearly non-responsive or stale content”, then ranks what is left in stages. Pages are split into “self-contained spans”, each ranked on its own. The post is about the Search API, though a caption says Perplexity today runs on the same infrastructure. It names stages, not factors.

What does Perplexity document, and what does nobody outside know?

Read this table as of November 4, 2026. “No” means we found it on none of the Perplexity pages we read, not that the company has no such rule.

The closest thing to a ranking statement sits on a shopping page. “What is Instant Buy?” (last modified October 26, 2026) says: “Just as our answer engine ranks sources based on authority and relevance, we rank product listings using similar criteria.” Perplexity defines neither word. So when you meet a list of “Perplexity ranking factors”, ask what it rests on: we found no such list from Perplexity.

Outside studies can only watch what comes out. An Ahrefs study dated August 11, 2025 ran 15,000 long-tail queries through Google and Bing, then asked the same questions of ChatGPT, Gemini, Copilot and Perplexity. It found 28.6% of the URLs Perplexity cited in Google’s top 10 for that query, against around 8% for the others. Even so, most of Perplexity’s cited URLs were outside that top 10. The study covers long-tail queries only, it is more than a year old, Ahrefs calls its explanations theories, and Ahrefs sells SEO tools. It shows an overlap, not a cause.

What happens if you block PerplexityBot or Perplexity-User?

PerplexityBot builds search results. Perplexity-User visits a page while a person waits for an answer. And robots.txt only asks a bot to stay out, while a firewall refuses it.

So a block on Perplexity-User in robots.txt may change little, while the same block at the firewall can stop the visit. If you want to be a source, allow both names at both layers.

Two Perplexity pages read differently here. The help page on robots.txt (last modified September 14, 2026) says users could once have it summarize a blocked URL, but “this feature has been disabled to prevent misuse”. The crawler page says Perplexity-User generally ignores robots.txt. Neither explains how the two fit, so read Perplexity’s robots.txt commitment as covering PerplexityBot only.

Three checks confirm access:

• Read robots.txt. Open yourdomain.com/robots.txt. Look for “Disallow: /” under PerplexityBot, or under “User-agent: *” when PerplexityBot has no group of its own. Perplexity says a change may take up to 24 hours to register. The lines to copy are in our guide to AI crawlers.
• Check the firewall. Perplexity says that behind a firewall “you may need to explicitly allow” its bots. For Cloudflare, its docs give a Skip rule that matches each bot’s name to that bot’s own addresses, placed before your blocking rules, and warn that Bot Fight Mode “can still challenge Perplexity bots that match this rule”. Cloudflare’s own reference lists PerplexityBot under AI Search and Perplexity-User under AI Assistant, so a setting built on categories can let one through and stop the other. With any other host, ask support about both names.
• Read your logs. Search your server or CDN logs for each name separately. A 200 means the bot received the page; 403 or 429 means it was turned away. A name alone proves nothing, so compare the address with perplexity.ai/perplexitybot.json or perplexity.ai/perplexity-user.json. These are “updated regularly”, so a rule typed once goes out of date: ask your developer to automate it, as Perplexity advises.

How do you read Perplexity’s citations for your own buyer questions?

This shows which pages were used for one answer on one day. It does not show why. Do it in this order:

• Write 10 questions a buyer would ask, without your name in them. Ten is our suggestion for a first look. Add your city where location matters.
• Ask in a private window, in Search, on Web. Note your plan, because the help pages describe different source counts. Perplexity also says past searches, interactions and saved preferences refine its product recommendations, so your usual account may see a different list.
• Open every numbered citation. Copy the full address of each page, not only the site name. Select or hover over a source to see any label: some carry a shield icon with Government, Academic or Trusted.
• Read how it broke the question down. Perplexity says Pro Search shows “how the AI broke down your question and approached the research”. Those narrower searches are questions your page could answer.

Then sort what you found:

A site that keeps coming back across your questions is one Perplexity leans on for your topic. That is our reading, because Perplexity publishes no trust score for a page. Most domains carry no label, and its help page says that “does not mean a site is low quality”.

What should you write down?

This is our suggestion, not a Perplexity feature: one row per cited page, the same 10 questions, once a month. The sample row is invented.

After three rounds, count how many questions each page appears for, and start with the pages that come up most. Compare rows only when the plan and mode match, and judge nothing on one round: answers vary from try to try.

Which details give Perplexity something to quote?

Perplexity gives a few clues, none of them a promise. Habits that apply to any assistant are in our guide to getting cited by ChatGPT.

• Answer first, in a passage that stands alone. If pages are split into “self-contained spans”, a passage that needs the paragraph before it is a weaker piece (our reading). Put the answer in the first sentence under the heading, and name the thing instead of writing “it”.
• Dates. The research post says the pipeline removes “stale” content, without saying how it judges that. An honest “updated” date is the plain way to show a page is current.
• Names. The source review behind the labels asks questions such as “Does the site correct its mistakes?” and “Does it say who wrote each piece?”. The page does not say these affect what gets cited. They are still the only site-quality questions we found from Perplexity, and they cost little: a byline and a corrections note.
• Prices and specifics. The Instant Buy page says: “Our answer engine is more likely to recommend merchants who provide deeper product details such as availability, reviews, pricing, and specifications.” That sentence is about products. For a service business, stated price ranges, service areas and lead times are the nearest match, which is our inference and not Perplexity’s rule.

What did Cloudflare and Perplexity each say about its crawling?

On August 4, 2025, Cloudflare published a post saying it had seen “stealth crawling behavior” from Perplexity. It says it created new domains, closed them in robots.txt and blocked Perplexity’s declared bots with firewall rules, and that Perplexity was “still providing detailed information” about their content. It lists 20 to 25 million daily requests from the declared Perplexity-User and 3 to 6 million from a generic browser identity. It says it de-listed Perplexity as a verified bot and added rules to block that traffic, and that once blocked, answers used other sources and “were less specific and lacked details from the original content”.

Perplexity replied the same day in “Agents or bots? Making sense of AI on the open web”. It says the 20 to 25 million are user requests and not scrapers. It says the 3 to 6 million came from BrowserBase, a third-party cloud browser service it “only occasionally uses” (“less than 45,000 daily requests”), and that Cloudflare misattributed them. It says its user-driven agents fetch a page only when a person asks, and “do not store the information or train with it”.

We cannot settle this from outside, and we take no side. Two practical points hold either way:

• If you want Perplexity to read you, do not assume that a setting for “verified bots” covers it. That post is from 2025, so look up today’s status in the bots directory on Cloudflare Radar.
• If you need a page kept private, robots.txt is a request and not a lock. Use a login.

Can you pay or join a program to be cited?

We found no published way to buy a citation. None of the program posts we read says a partner is cited more often, and none rules it out. We could not confirm current terms for any of them.

Two help pages speak to money directly. The Instant Buy page says advertisers “cannot pay to list their products” in its related products section. The label page says “Perplexity’s partnerships, payments, and other business arrangements do not affect a site’s label.”

Where the figures come from

• Perplexity’s help and docs pages, read November 4, 2026: “Perplexity Crawlers” (no date shown); “How does Perplexity work?”, “What is Pro Search?” and “What is Instant Buy?” (last modified October 26, 2026); “Understanding source labels” (October 6, 2026); “How does Perplexity follow robots.txt?” (September 14, 2026).
• Perplexity’s posts: the research post of September 25, 2025 and the reply of August 4, 2025. Their figures are its own.
• Cloudflare: its post of August 4, 2025, with its own test and figures, and its AI Crawl Control bot reference.
• Ahrefs: “Only 12% of AI Cited URLs Rank in Google’s Top 10 for the Original Prompt”, August 11, 2025.

What the MWY checker does and does not tell you about Perplexity

MWY’s free AI visibility checker covers access. It reads your robots.txt and tests both PerplexityBot and Perplexity-User against it. It also asks for your home page under PerplexityBot’s name to spot a firewall refusal, and marks that as a signal, because the real bot comes from its own addresses. The free result gives your score, the score of each group and the five items that matter most.

It does not test Perplexity-User at the firewall, read your logs or compare addresses with Perplexity’s lists. And the free check tells you nothing about Perplexity’s answers: not whether it cites you, and not which sources it uses for your topic. For that, ask it your buyer questions.

Pro, which opens soon, adds the answers. Every week, 20 buyer questions go to ChatGPT, Gemini, Claude and Perplexity through their APIs with web search on, and the report shows which answers name your site and which pages they cite. Answers through an API can differ from what one person sees in the app. Nothing on this page, and nothing MWY does, makes Perplexity cite a business.