How Perplexity Chooses Sources

Updated 2026-10-24 · By the GEO Checker team

Perplexity is an AI answer engine that searches the web for every question, reads several pages and writes an answer with numbered citations. Perplexity has not published a ranking formula, so nobody can tell you exactly why one page is cited over another. What is documented is how its crawlers work, and what is observable is which kinds of pages tend to be cited. This guide combines both into a practical checklist.

How a Perplexity answer is built

At a high level, Perplexity (1) interprets the question, (2) retrieves candidate pages from its own index and live web search, (3) reads the most relevant passages, and (4) writes an answer that cites the pages it relied on. Your page has to survive each stage: it must be in the index or fetchable, it must look relevant for the query, and it must contain a passage that answers the question clearly enough to be quoted.

Perplexity documents two user-agents. PerplexityBot crawls the web to surface and link websites in Perplexity search results; Perplexity says it respects robots.txt and publishes its IP ranges. Perplexity-User fetches pages when a user’s question requires visiting a specific URL; because that request is user-initiated, Perplexity says it generally ignores robots.txt rules. If you block PerplexityBot, you should expect to drop out of Perplexity’s index, which makes citations much less likely.

What tends to get cited

• Direct answers near the top. Pages that state the answer in the first paragraph or a clearly labeled section are easy to quote.
• Freshness. For time-sensitive topics, recently updated pages with visible dates are favored.
• Specific facts. Numbers, named entities, steps and tables give the model something concrete to cite.
• Reputation. Established domains, documentation, forums such as Reddit and well-known reviews appear frequently, so being mentioned there also helps.
• Accessible HTML. Content rendered on the server and not hidden behind scripts, logins or bot challenges.

Step by step

1. Allow PerplexityBot in robots.txt. Make sure no robots.txt group disallows PerplexityBot for your public pages. The AI crawler checker shows the exact rule that applies.
2. Make sure your firewall lets it through. Cloudflare and other CDNs can block or challenge PerplexityBot regardless of robots.txt. Run the free AI visibility check, which compares PerplexityBot’s response with a browser’s, and allow Perplexity’s published IP ranges if your WAF verifies bots.
3. Answer the question in the first 100 words. Rewrite key pages so the opening paragraph contains a direct, self-contained answer, followed by details. Use descriptive headings that match how people phrase questions.
4. Add facts, tables and dates. Include concrete numbers, comparison tables, step lists and a visible “last updated” date. Keep facts consistent across your pages, structured data and llms.txt.
5. Use structured data. Add Organization, Article, FAQPage, HowTo and Product JSON-LD where relevant. Stores should run the AI shopping agent readiness check so product details, prices and policies are machine-readable.
6. Build presence on sources Perplexity already cites. Contribute genuinely useful answers in communities, get listed in respected directories and earn reviews and comparison mentions. When Perplexity cites a forum thread or review that names you, your brand appears in the answer even if your own page is not cited.
7. Track crawls and citations. Use AI crawler analytics to confirm PerplexityBot fetches your important pages, and check a fixed set of questions in Perplexity weekly to see whether you are cited.

Example: a passage that is easy for Perplexity to cite

Question: “What is the return window at Northwind Coffee?”

Hard to cite: a returns page that opens with a brand story, puts the policy in an image, and mentions “30 days” only in the third paragraph.

Easy to cite: “Northwind Coffee accepts returns of unopened bags within 30 days of delivery. Refunds go to the original payment method within 5 business days.” Followed by a short FAQ, a MerchantReturnPolicy in JSON-LD and a visible updated date.

Perplexity vs ChatGPT Search: what is different

The practical takeaway: the same technical foundations work for both. If PerplexityBot and OAI-SearchBot can fetch fast, server-rendered pages with clear answers, you have done most of what is in your control. For the ChatGPT side, see how to get ChatGPT to recommend your website.

What does not work

• Keyword stuffing. Answer engines read passages; repeating a phrase does not make a passage more useful.
• Hidden text for AI only. Content shown to bots but not users can be treated as cloaking by search engines.
• Fake reviews or ratings in schema. They violate structured data guidelines and undermine trust.
• Mass-produced thin pages. Dozens of near-identical pages rarely earn citations and can hurt how search engines see the site.