Perplexity fetches sources live for every question. There is no fixed ranked index
Perplexity chooses sources by running a live retrieval pass for each question: it fetches a set of candidate pages at that moment, then narrows them down by relevance, freshness, and how cleanly a fact can be lifted from the page, before citing a small number in the answer. There is no static leaderboard and no paid placement.
Short answer
Perplexity does not consult a pre-built ranking the way a classic search engine does. For each question it retrieves a set of candidate pages in real time, evaluates them against the specific query, and cites the ones that are relevant, current, and clear enough to state a fact from with confidence. The set of sources can differ between two people asking almost the same question minutes apart.
Retrieval happens per question, not once in advance
Where a traditional search engine ranks the whole web against a query using an index built ahead of time, Perplexity runs its retrieval step live. Each question triggers a fresh search across its own index and partner sources, pulling in a working set of candidate pages before any citation is chosen.
That live step is why the sources Perplexity cites for a topic can shift from one week to the next even when nothing about the underlying web page changed: the retrieval is re-run every time, not looked up from a fixed table.
What narrows the candidate set down to a citation
Relevance to the exact question comes first. A page that is broadly authoritative on a topic but does not address the specific thing asked is a weaker candidate than a smaller page that answers it directly.
Beyond relevance, two qualities repeatedly show up in how Perplexity and similar systems describe their own approach: recency, because a live retrieval system can and does favour content that has been published or updated more recently over an older page saying the same thing, and extractability, meaning a claim stated in one clear sentence is easier for the system to lift and attribute correctly than the same fact buried inside a long, qualified paragraph.
A source is not on a list waiting to be picked. It is fetched, read and judged again every time someone asks.
Two different crawlers, doing two different jobs
Perplexity operates more than one crawler, and the distinction matters for anyone checking server logs. PerplexityBot is the general-purpose crawler that builds Perplexity's own search index over time. Perplexity-User is a separate crawler that fetches pages in real time while it is actively answering a specific question, which is why a page can show a Perplexity-User request without ever having been indexed by PerplexityBot first.
Perplexity's own help documentation states that PerplexityBot respects robots.txt and will not index the full text of a page a site has disallowed, though it may still surface the domain, headline and a brief summary. Changes to robots.txt can take up to 24 hours to take effect on Perplexity's side.
The robots.txt question is not fully settled
In August 2025, Cloudflare published a report accusing Perplexity of continuing to crawl sites after being blocked, by repeatedly changing its declared crawler identity and network source, and in one case using a generic user agent designed to look like a Chrome browser. Cloudflare said it observed millions of these requests daily across tens of thousands of domains and de-listed Perplexity as a verified bot as a result.
Perplexity disputed the characterisation and said the specific bot named in Cloudflare's evidence was not one of its own. Neither side's account has been independently settled in public since. The practical takeaway is not that robots.txt is worthless, but that it should not be treated as a guaranteed enforcement mechanism against every crawler that might read a page.
What this means for a business page
Because retrieval is live rather than fixed, a page that is easy to fetch, clearly structured, and states its facts in short, direct sentences has a real advantage over one that is technically correct but buried in qualifying language. This is the same discipline answer engine optimization is built around, applied to a system that re-evaluates the page on every question rather than once.
Keeping pages current matters more here than for a static index: a page that has not been reviewed in a year is competing against fresher pages saying the same thing, on a system that is checking that freshness at the moment of the question.
Common questions
Does Perplexity use a fixed ranking like a search engine?
No. Perplexity runs a live retrieval pass for each question rather than consulting a static ranked index. It fetches a set of candidate pages at the moment of the query, then narrows that set down to the handful it actually cites in the answer.
Can a business pay to be cited by Perplexity?
No. There is no advertising or paid placement mechanism for organic citations. A source gets cited because it was retrievable, relevant and clear enough to lift from, not because anyone paid for a slot.
Is PerplexityBot the same crawler that answers my question?
No. PerplexityBot is the indexing crawler that builds Perplexity's general search database. Perplexity-User is a separate crawler that fetches pages in real time while answering a specific question. A site can be indexed by one and still be fetched live by the other.
Does blocking PerplexityBot in robots.txt stop Perplexity from citing a page?
It stops the full text from being indexed, according to Perplexity's own help documentation. A blocked page may still have its domain, headline and a brief factual summary surfaced, just not its full content, and changes to robots.txt can take up to 24 hours to take effect on Perplexity's side.
Why did Cloudflare accuse Perplexity of ignoring robots.txt?
In August 2025, Cloudflare said Perplexity kept crawling sites after being blocked by rotating its declared crawler identity and network addresses, and in some cases impersonating a Chrome browser. Perplexity disputed the characterisation. Both the accusation and the denial are on the public record, and a business relying on robots.txt alone should know the compliance question has been contested.
AEO services
Building visibility across every answer engine, not just one.
How does ChatGPT choose which businesses to recommend
The same question asked of a different engine, with a different retrieval model.
Does ChatGPT use Google search results
Why answer engines often run on completely separate retrieval infrastructure.
What is an answer engine
The category Perplexity belongs to, and how the category works.
Find out whether Perplexity can actually fetch and cite your site today.