Back to blog

Engine Optimization

How Perplexity Decides Which Sources to Cite in Its Answers

Perplexity's source selection explained. The technical and content signals it weights, and concrete steps to earn more Perplexity citations.

Louis Foweraker · ·

Perplexity leans on pages that already rank in search more than any other AI assistant: in Ahrefs' analysis of 15,000 prompts, 28.6% of the URLs Perplexity cited ranked in Google's top 10 for the same query, against an average of 12% for ChatGPT, Gemini and Copilot. That is the clearest public signal of how Perplexity decides which sources to cite. It searches the web for every question, pulls candidate pages from its own index, and cites the passages that answer the question most directly. Perplexity does not publish its ranking factors, so what follows sticks to what it documents and what controlled tests have measured.

What Perplexity documents about its sources

Perplexity runs two separate agents, and its crawler documentation explains the difference. PerplexityBot builds the index that search results are drawn from, and Perplexity recommends allowing it in robots.txt for a site to appear in those results. Perplexity-User fetches a page when a person's question requires it, and because a user triggered the request, it generally ignores robots.txt rules.

The practical consequence is simple. A site that disallows PerplexityBot removes its pages from the pool Perplexity normally cites from, while still receiving occasional live fetches when someone asks about a specific page. Our guide to robots.txt and AI crawlers covers the directives each bot obeys.

Why Google rankings matter more on Perplexity

Perplexity behaves like a search engine that writes its results as prose. Because it retrieves fresh results for each question, the pages that search systems already rank well are the ones most likely to land in its candidate pool, which is what the Ahrefs overlap shows. SEO groundwork such as indexability, clear titles and internal links pays off more directly here than on ChatGPT, which cites pages ranking in Google's top 10 far less often.

Ranking is still not the same as being cited. Most of Perplexity's citations in the Ahrefs data came from outside Google's top 10, and a page can be retrieved without any of its passages making it into the answer, a distinction covered in retrieval vs citation in AI search. For the contrast with OpenAI's engine, see how ChatGPT chooses its sources.

What content Perplexity is more likely to quote

The only controlled test run on Perplexity itself comes from the original GEO study. Researchers fed 200 queries and their source documents to Perplexity and rewrote the sources with different methods. Adding statistics and adding quotations improved how much and how prominently a source was used: the best methods raised position-adjusted word count by 22% and subjective impression by 37%. Keyword stuffing scored below the unedited baseline on position-adjusted word count.

Perplexity quotes passages, not pages. A sentence that states a specific fact with its source, near the top of a section, is the unit that gets cited.

In practice that means opening each section with the answer, keeping figures next to their named source, and writing passages that make sense when lifted on their own. Our guides to answer-first writing and to how AI engines chunk content before citing it show how to structure those passages, and statistics addition explains why numbers help.

The crawling dispute and what it means for site owners

Perplexity's relationship with publishers has not been frictionless. In August 2025, Cloudflare reported that when its declared crawlers were blocked, Perplexity appeared to use undeclared crawlers presenting as a regular browser to reach the same content, and it removed Perplexity from its list of verified bots. Perplexity disputed the findings.

For a brand that wants to be cited, the takeaway is to decide access deliberately rather than inherit it from a CDN default, as described in what Cloudflare's AI crawler blocking does to visibility. For a site that wants out, robots.txt alone may not be enough, and the fakes and look-alikes covered in how to verify AI crawlers need checking against Perplexity's published IP ranges.

How to check whether Perplexity cites you

Perplexity shows numbered sources under every answer, which makes it the easiest engine to audit by hand. Run your buyer prompts several times, note whether your brand is named and which domains are cited, and compare the result with competitors over time; the method is detailed in how to track brand mentions in ChatGPT, Gemini and Perplexity. The share of answers that cite one of your pages is your citation rate on Perplexity.

Your server logs add a view no answer shows. Each Perplexity-User request marks a page fetched live because someone asked a question that needed it, so a rising count on a page tells you which content buyers are pulling into their answers, even when they never click through. Crawler log analysis explains how to separate those visits from PerplexityBot's indexing crawl.

Frequently asked questions

How does Perplexity decide which sources to cite?

Perplexity searches the web for every question, retrieves candidate pages from its index and cites the passages that answer the question most directly. It does not publish its ranking factors. Public data shows it favors pages that already rank well in search: in an Ahrefs study, 28.6% of its cited URLs ranked in Google's top 10.

Should I allow PerplexityBot in robots.txt?

Yes, if you want to be cited. Perplexity's documentation recommends allowing PerplexityBot for a site to appear in its search results, since that crawler builds the index Perplexity cites from. Its other agent, Perplexity-User, fetches pages when a user's question requires it and generally ignores robots.txt.

Does ranking on Google help with Perplexity citations?

It helps more than on other AI assistants. Ahrefs found that 28.6% of Perplexity's cited URLs ranked in Google's top 10 for the same query, against an average of 12% for ChatGPT, Gemini and Copilot. Most Perplexity citations still come from pages outside the top 10, so ranking is not a guarantee.

What kind of content does Perplexity cite most?

Content that states facts directly, with statistics and named sources, is more likely to be used. In a controlled test on Perplexity from the original GEO research, adding statistics and quotations raised subjective impression scores by up to 37%, while keyword stuffing reduced how much of the text was used compared with the unedited version.

How can I see whether Perplexity cites my website?

Run the questions your buyers ask in Perplexity several times and check the numbered sources under each answer. Record whether your brand is named and which domains are cited, and compare the results with competitors over time. Server logs showing Perplexity-User requests also reveal which pages are fetched live for users' questions.

Related articles

Brand Visibility

How to track brand mentions in Microsoft Copilot, starting with Bing's citation report (2027)

Bing Webmaster Tools tells you, for free, how often Copilot cites your pages. This guide shows what that report counts, what it leaves out (brands named without a link, answers given at work) and how to fill the gaps with prompt checks and GA4.

Louis Foweraker ·

Brand Visibility

Google AI Overview tracking: find the keywords that trigger one, then the brands it names (2026)

Tracking Google AI Overviews starts with keywords, not prompts: an overview appears on about a quarter of US queries, and far more often on long informational ones. This guide shows how to record the trigger, the linked sources and the brands named without a link, and what Search Console and rank trackers leave out.

Louis Foweraker ·

Brand Visibility

Best AI visibility tools for small businesses in 2026: 11 affordable options compared

Many AI visibility tools start at $95 a month or more. Here are the 11 best options for a small business on a budget, with who each one suits, its real monthly cost, its pros and cons, what its users report and the free checks to run before you pay for anything.

Louis Foweraker ·

Brand Visibility

AI visibility tracking tools under €50 a month: 2026 comparison

At least six plans track how AI assistants mention a brand for €50 a month or less. We checked each vendor's pricing page on 9 November 2026 and compared what the money buys: prompts, AI engines and how often they are checked.

Louis Foweraker ·

Optimize your presence across Generative AI and Answer Engines.

Track your AI visibility · Get a demo