Perplexity SEO: how Perplexity picks sources

Perplexity SEO is the work of making your pages easy for Perplexity to crawl, retrieve and quote. It is a narrower job than it sounds, because Perplexity has published more about how it picks sources than most answer engines. Its engineering write-up on the Perplexity search stack describes the crawler, the index, how documents are split into passages and how those passages are ranked (Perplexity, September 2025).

This explainer follows that pipeline from crawl to citation and marks where you have leverage and where you do not.

How does Perplexity find your pages?

Perplexity runs its own crawler and index. Perplexity says its index “tracks over 200 billion unique URLs”, that PerplexityBot “complies with explicit limits set forth in robots.txt files”, and that a machine-learning model decides which URLs to index or refresh, weighted by how important a URL is and how often it is likely to change. Refreshing known pages competes with discovering new ones, so not every page is revisited often.

What that means for you:

• PerplexityBot must be allowed in robots.txt and must not be blocked by your CDN. A page outside the index cannot be cited.
• Important pages need clear signals of importance: internal links from your main navigation and hub pages, and inclusion in your XML sitemap.
• Real updates beat cosmetic ones. Our view: if the refresh model learns that a page changes meaningfully, it has reason to revisit it. Changing a date without changing the content gives it no such reason.

How does Perplexity choose sources for an answer?

Perplexity splits documents into self-contained passages, retrieves candidates with both keyword and semantic matching, then reranks them in several stages, ending with cross-encoder models that score each passage against the query. Perplexity says it does not choose between lexical and semantic retrieval but merges both into one candidate set, and trains its ranking models on AI evaluations combined with live human feedback.

Three details from the write-up matter most:

1. Passages, not pages. Perplexity says it decomposes documents into “self-contained spans, each of which can be individually retrieved and ranked”. A page is cited because one of its sections answered the question well.
2. Exact wording still counts. Lexical retrieval means the specific terms people use (product names, model numbers, error codes) still help a passage get retrieved.
3. Structure is preserved. Perplexity grades its parsing on whether it keeps the original content structure and layout. Headings, lists and tables that are real HTML carry their meaning into the index.

This is a form of retrieval-augmented generation: the model writes an answer from passages the search system hands it, and the citations point to those passages.

Which parts of the pipeline can you influence?

Stage	What Perplexity says	What you can do
Crawl	PerplexityBot follows robots.txt	Allow PerplexityBot; check that your firewall does not block it
Index and refresh	An ML model predicts which URLs to index or refresh	Internal links, sitemap, substantive updates
Parse	Documents become self-contained spans; layout is preserved	Real headings, lists and tables; one topic per section
Retrieve	Lexical and semantic results are merged	Use the exact terms your audience searches, alongside plain explanations
Rank	Multi-stage ranking ending in cross-encoder rerankers	Answer the question directly in the first lines of the section
Cite	The answer links the passages it used	Make each passage accurate and attributable: numbers with sources, named experts

The last three rows are where most of the work lives, and they reward the same writing habits: a question-shaped heading, a direct answer, then the supporting detail.

Our view on sections: Google’s guidance says you do not need to “chunk” content into tiny pieces, and that holds here too. The aim is not short fragments but sections that make sense on their own. A section that opens with “As mentioned above” loses its meaning when it is lifted out as a single span.

PerplexityBot or Perplexity-User: which one matters?

Both, for different reasons. Perplexity says PerplexityBot surfaces and links websites in search results and is not used to crawl content for training AI foundation models. Perplexity-User visits a page when a person asks a question that needs it; Perplexity says this fetcher “generally ignores robots.txt rules” because a user requested the fetch (Perplexity crawler documentation).

Agent	Job	Honors robots.txt	Effect of blocking
PerplexityBot	Builds the search index	Yes	Your pages drop out of the index Perplexity answers from
Perplexity-User	Live fetch for a user’s question	Generally no	Needs a firewall rule; live answers about your page lose access to it

Perplexity publishes IP lists for both agents as JSON files and recommends automating updates to firewall rules. Changes to robots.txt can take up to 24 hours to apply. If you want to block Perplexity entirely, robots.txt alone will not do it; if you want to be cited, allowing PerplexityBot is the one step that is not optional. Copy-paste policies are in robots.txt for AI bots, and the robots.txt tester shows how your current file treats each Perplexity agent.

Does GEO research apply to Perplexity?

Partly. The KDD 2024 GEO paper tested its content edits on Perplexity.ai as well as on the authors’ own benchmark engine (Aggarwal et al.). On Perplexity, adding quotations improved the paper’s main visibility metric by 22% over unedited sources, adding statistics improved its secondary metric by up to 37%, and keyword stuffing performed 10% worse than doing nothing.

Research shows, with two caveats. The tests ran on a fixed benchmark of queries and sources, not on live competition for real prompts. And the Perplexity of 2023 is not the Perplexity described in the 2025 engineering post, which uses a different index and ranking stack. The direction of the findings (specific, sourced, quotable passages help; stuffing hurts) is consistent with how the current pipeline is described. The size of the effect should not be quoted as a forecast.

How do you check whether Perplexity cites you?

1. Confirm access. Run the AI visibility checker to see whether your robots.txt allows PerplexityBot and whether a request using its user agent receives your page.
2. Build a prompt set. Write 20 to 30 questions your audience asks, run them in Perplexity, and record whether your pages appear in the sources list and which competitors appear instead.
3. Track referrals. Filter analytics for perplexity.ai. Clicks understate visibility, because many people read the answer without opening sources.
4. Repeat monthly. Compare the share of prompts where you are cited, not individual answers, which vary from run to run.

If you cover several answer engines, the same prompt set works across all of them. The AI crawlers guide lists every agent you may want to verify.

What to do next

Start with access, because nothing else matters until PerplexityBot can fetch your pages. Then pick the five pages you most want cited and rewrite their key sections so each one opens with a direct answer, uses the exact terms your audience searches, and backs claims with a source. For a parallel walkthrough on OpenAI’s side, see how to get cited by ChatGPT.

Frequently asked questions

Does Perplexity respect robots.txt?

PerplexityBot, the crawler that builds Perplexity's search index, complies with robots.txt according to Perplexity. Perplexity-User, which fetches a page when a person asks a question, generally ignores robots.txt because the fetch is user-initiated. To stop Perplexity-User you need a firewall rule, using the IP ranges Perplexity publishes.

Is Perplexity SEO different from Google SEO?

The foundations are the same: crawlable pages, clear text, relevant answers. The differences are in emphasis. Perplexity retrieves and ranks self-contained passages rather than whole pages, and it maintains its own index, so you also need PerplexityBot to be allowed and able to fetch your pages.

Does Perplexity use Google or Bing results?

Perplexity says it runs its own search index, which tracks over 200 billion URLs, and its own crawler, PerplexityBot. It has not said that it relies on Google or Bing for answers. Pages that are blocked from PerplexityBot cannot enter that index.

How do I see traffic from Perplexity?

Look for referrals from perplexity.ai in your analytics. Not every citation produces a click, so pair referral data with a fixed set of prompts you run each month and record whether your pages appear in the sources list.