Perplexity SEO: How Perplexity Chooses Its Sources

Decide the access rule for each Perplexity agent before you change robots.txt. The two agents follow different rules, so one Disallow line does not control both.

Perplexity recommends that you allow PerplexityBot in robots.txt and permit requests from its published IP ranges, so your site can appear in results.

A Disallow rule for PerplexityBot does not govern Perplexity-User. Perplexity says the user-triggered agent generally ignores robots.txt, so do not count on robots.txt to stop a fetch started by a question.

Robots.txt changes can take up to 24 hours to reach Perplexity. Each setting works independently. Test a rule change after that window, not right after you save it.

Perplexity publishes user agent strings with a version number: PerplexityBot/1.0 and Perplexity-User/1.0. Match on the agent name, not the full string, because the version can change.

Firewall rules need the right order, and the order decides whether an allow rule works. Perplexity says Cloudflare's Bot Fight Mode cannot be skipped with WAF custom rules. If that mode is on, an allow rule alone does not stop the challenge.

For AWS WAF, Perplexity's guide describes one IP set per bot, plus an Allow rule per bot that matches both the IP set and the User-Agent. Give those Allow rules lower priority numbers than your blocking rules, so the WAF evaluates them first.

Each bot publishes its own IP list as a JSON file. Match both the User-Agent and the IP address before you accept a hit.

Use the most current IP ranges from the official JSON endpoints. Lists that you saved last year may no longer match.

Allow PerplexityBot if you want to appear in Perplexity results. Treat Perplexity-User as a fetch that robots.txt generally does not stop. Before you change any access rule, run the crawl and indexing checks in the SEO guide for practitioners.

Plan for passage-level selection, not only for page-level ranking. Perplexity says it retrieves and scores results at both the document and sub-document levels.

Perplexity's Search API post says its index splits documents into units that are scored one at a time against the query. The API then returns documents with their most relevant snippets already ranked.

Perplexity says its search index tracks over 200 billion unique URLs, with capacity for many hundreds of billions more.

Perplexity says it built its own index and crawler for completeness, freshness and speed, and for fine-grained content understanding. Its retrieval runs lexical and semantic queries together, then merges the results into one candidate set.

The ranking runs in stages. Perplexity says prefilters remove clearly non-responsive or stale content, and lexical and embedding scorers narrow the set. Cross-encoder rerankers make the final selection.

Perplexity says its indexer keeps documents from authoritative domains, documents on undercovered topics, and other categories with outsize potential to enrich the index. It gives no weights for that signal, so do not size your effort from it.

Perplexity says the internet is too large to index every URL on a frequent, recurring basis. A machine-learning model decides which URLs to index, and when, based on importance and likely update frequency.

Update frequency is one input the indexing model weighs, so plan updates for the pages you want revisited.

Perplexity says its parsing rules vary by site type. A site with well-structured lists or tables may get more formulaic extraction than a site with freeform writing. If a page is a list or a comparison, format it as one, so the extraction has clear units to read.

The published material covers the index and the retrieval stages. It gives no weights for which sources an answer cites. Treat any vendor's citation weights as opinion, because none of the Perplexity pages cited here gives one.

Perplexity's help center says each answer includes numbered citations that link to the original sources. Run your own queries and record which numbered links point to your pages.

Mode changes the source count. Perplexity says Deep Research runs many searches on its own and reads hundreds of sources. A single source count does not describe every answer.

For the tactics that apply across AI search, read the GEO and AEO guide. The GEO vs SEO page explains where that work differs from classic SEO.

Do not plan a traffic or payment gain from Perplexity's publisher programs. Both name publishers as their partners, so the work that helps a marketing site is crawl access, passage quality and measurement.

Perplexity launched its Publishers' Program on July 30, 2024. Its first partners were TIME, Der Spiegel, Fortune, Entrepreneur, The Texas Tribune and WordPress.com.

Perplexity says it pays a share when it earns revenue from an interaction where a publisher's content is referenced. Partners also get free access to its online LLM APIs and developer support.

Perplexity announced Comet Plus on August 25, 2025. It says the program pays publishers for three types of traffic: human visits, search citations and agent actions.

Perplexity says it distributes all revenue to participating publishers, minus a small portion for compute costs. Its post gives no percentage and no dollar figure.

Digiday reports, quoting Perplexity's head of publisher partnerships, that Perplexity keeps 20% of the pooled subscription revenue. The other 80% goes to publisher program participants. Digiday also reports a $42.5 million pool.

Digiday says Comet Plus pays for direct visits through Comet, for citations in answers on Comet, and for tasks that Comet's assistant completes. Those payments do not depend on citations in Perplexity.ai web answers alone.

Digiday quotes the same executive saying publishers could stand to make millions from the program. Neither Perplexity's post nor Digiday gives a payout total.

The named partners are news and media companies, plus WordPress.com. Neither program names a brand or a marketing site as a partner, so do not plan a traffic gain around one.

Seer Interactive studied 7,683 dated pages cited across ChatGPT, Gemini and Perplexity, from March to June 2026. Of Perplexity's cited pages, 65% were updated within the last year, and 83% within two years.

Seer reads its data as showing that refreshing older pages is what earns recency. Its fresh-from-old share for Perplexity, meaning pages updated in the last year but first published two or more years ago, is 28%.

Seer covers four brands in four categories, so read these shares as one sample, not as a rule for every site.

A preprint on arXiv analyzed 602 controlled prompts across ChatGPT, Google AI Overview and Gemini, and Perplexity, with 21,143 valid citations. Its authors report that Perplexity and Google cite more sources per answer on average than ChatGPT does.

The same preprint finds that high-influence pages tend to be longer, more structured and richer in extractable evidence, such as definitions, numerical facts, comparisons and procedural steps. The preprint has not been peer reviewed.

A second arXiv study analyzed 158,099 answers and 468,037 citations from ChatGPT, Gemini and Perplexity, over two 90-day collections. Within one site, pages whose title and paragraphs match the questions people ask get cited more.

The same study finds that FAQ blocks, structured data, summary boxes and page speed appear effective only when compared across websites. It covers three engines together, so read its findings as cross-engine, not as Perplexity-only.

Qwairy measured 21.87 citations per question for Perplexity and 7.92 for ChatGPT, across 118,101 AI answers in Q3 2025. Use that count with its sample and period, and treat it as one measure of a single quarter.

For the cross-platform view, read the platform citation behavior page. For how the ChatGPT side differs, read how to get cited by ChatGPT.

Use each study's design to decide what to test. Then measure your own citations for your own questions, because no shared figure predicts your results.

Measure two series separately: referral visits from Perplexity, and bot visits from its crawlers. Bot visits show crawl access. They do not show who clicked through.

Ahrefs separates crawl hits from human referral traffic. It says AI crawlers fetch and index pages, but they do not necessarily send human visitors.