How Perplexity chooses its sources, and what you can do to get cited

Quick summary

• Perplexity is a search engine that answers by citing sources. It searches the web in real time, extracts passages from pages and writes the answer with numbered links.
• It runs two agents: PerplexityBot, which crawls for its search engine, and Perplexity-User, which visits pages when a user asks. According to its documentation, the latter generally ignores robots.txt, because a person requested the fetch.
• Perplexity doesn't publish its ranking criteria. Anyone listing "the 7 Perplexity ranking factors" is guessing.
• What you can control: PerplexityBot can get in (through your firewall too), the answer is in your page's text, and you're present on the third-party sources it already cites.

What's documented

The crawlers

Perplexity's crawler page describes two agents:

• PerplexityBot is designed to surface and link websites in Perplexity's search results. Perplexity states it is not used to crawl content for AI foundation models. It recommends allowing it in robots.txt and allowing requests from its published IP ranges.
• Perplexity-User supports user actions: when someone asks a question, it may visit a page to answer accurately and link to it. It's not used for web crawling or training. And because a user requested the fetch, it generally ignores robots.txt rules.

One practical consequence few people consider: if you block PerplexityBot in robots.txt, you drop out of Perplexity's search index, even though Perplexity-User may still read a page when someone explicitly asks for it.

The firewall

The same documentation has a whole section on configuring web application firewalls (WAFs) in Cloudflare and AWS: rules that allow PerplexityBot and Perplexity-User by combining the user-agent with the IPs published in perplexitybot.json and perplexity-user.json.

If Perplexity bothers to document it, it's because it's a real problem: plenty of sites have a correct robots.txt and still have a CDN blocking "unknown" bots. You can't see that from robots.txt; you have to check firewall logs.

How it searches

Perplexity's API gives clues about how its search works. Its Search API returns "real-time, ranked web results", with controls for domains, languages, regions and extracted content. Each result includes a text snippet taken from the page, not just the title. Answers generated with search include citations to those sources.

Two things follow from that, without inventing more:

1. What counts is text that can be extracted. If your answer lives inside an image, a poorly tagged PDF or a component that only renders with JavaScript, it's harder for it to end up in a snippet.
2. Search is real time. A new page can start being cited as soon as it's indexed, without waiting for a model to be retrained.

What isn't documented

Perplexity doesn't publish how it orders results or why it cites one source over another. We haven't found an official document that says how much domain authority, date or page format weigh. So anything you read about "Perplexity ranking factors" is, at best, third-party observation.

What you do see when you measure regularly:

• Sources change between runs of the same question, as in ChatGPT or Gemini.
• Lots of third-party pages get cited: comparisons, forums, reviews, documentation. For "best X for Y" questions, your site competes with the roundups that talk about you and your competitors.
• Answers differ by language. The same question in English and Spanish can cite completely different sources.

Checklist for appearing in Perplexity

1. Access

• robots.txt allows PerplexityBot (and you don't have a blanket User-agent: * with Disallow: / catching it). Check it with the AI crawler checker.
• Your firewall or CDN doesn't block its IPs or user-agent. Check your block logs.
• The page returns a 200 and the content is in the HTML, not only after JavaScript runs.

2. Citable content

• Answer up top. If the question is "how much does X cost", the figure (with its date and source) goes in the first paragraph, not after four paragraphs of preamble.
• Paragraphs that stand on their own. An extracted passage without context should still say something useful.
• Verifiable data with sources. An engine that cites sources has reasons to prefer pages that cite theirs.
• Visible dates on content that goes stale: prices, comparisons, regulations.
• Tables and lists in HTML, not in images.

3. Presence off your site

Look at which sources Perplexity cites for your buyer questions. If they're third-party roundups or comparisons, the useful question is whether you could legitimately be in them: a genuine review, a correction to outdated information, a listing in a directory your industry really uses. More in why AI cites your competitors instead of you.

4. Measurement

Ask your buyer questions in Perplexity several times, in clean sessions, and record whether you're mentioned, whether you're cited and which URLs are cited. Clicks from Perplexity carry its domain as the referrer, so you can see them in Analytics. GA4 has an "AI Assistant" channel for assistants on its list (the documentation names ChatGPT, Gemini, DeepSeek, Copilot and Grok without publishing the full list), so it's worth filtering by source as well. We cover it in how to track ChatGPT traffic in GA4, which applies equally to Perplexity.

Mistakes we see often

• Blocking "all AI bots" with one CDN toggle. It usually catches PerplexityBot along with training crawlers, and you lose search visibility to block training.
• Testing only in English. If you sell in several markets, ask the same question in each language; the cited sources are often entirely different.
• Judging from one answer. A single run where you appear (or don't) tells you very little. Look at shares across repeated runs.
• Ignoring what gets cited instead. The third-party URLs Perplexity cites are the most actionable part of the data.

How Citable measures it

Citable measures Perplexity from the mid-tier plan up. It does so through Perplexity's model API with web search, not the consumer app, so answers may not match exactly what a user sees on perplexity.ai. We say so in the methodology because it's a real difference. Each question is asked three times a week, and mentions, citations, competitors and the third-party URLs cited instead of you are recorded.

FAQ

Does Perplexity use my content to train models?

According to its documentation, PerplexityBot isn't used to crawl content for training AI foundation models, and neither is Perplexity-User. Both serve search and answers.

Can I block Perplexity with robots.txt?

You can block PerplexityBot, which removes you from its search index. Perplexity-User, which visits pages at a user's request, generally ignores robots.txt according to Perplexity itself. If you want to appear in its answers, it's sensible to allow both.

Does Perplexity favour big domains?

Perplexity doesn't document it. In practice, both large domains and small, very specific pages get cited. What you can control is that your page is crawlable and answers the question better than the sources it cites today.

How long does Perplexity take to cite a new page?

There's no published timeline. Because it searches in real time, a page can be cited once PerplexityBot has crawled and indexed it. Measure before and after with repeated runs to know whether anything changed.

Conclusion

We know quite a lot about how Perplexity gets into your site and almost nothing about how it ranks. So start with what you can verify: PerplexityBot gets past robots.txt and your firewall, your answer is in extractable text, and you're present on the sources it already cites. And measure with repetitions before concluding anything.