Skip to content
HowFindableScan
Back to blogGuide

How Perplexity picks the sources it links

September 17, 2026·4 min read

Perplexity shows its work. Every answer comes with numbered links along the side, and those numbers are pages it went out and read while the question was being typed. Click one and you land on somebody's site, which is the part that matters to you.

That makes Perplexity the easiest of the big answer engines to reason about. There is no guessing about what a model absorbed during training two years ago. If its crawler could reach your page at the moment someone asked, you were in the running. If it could not, you were not.

How Perplexity builds an answer from live sourcesA question triggers a live search, PerplexityBot reads the pages it can reach, and the answer lists them as numbered links.Someone asksa questionPerplexityBotreads pages it canreach, right nowAnswer, withyour link in it[1] [2] [3]Nothing here is remembered from training: it is fetched at the moment of the questionIf the crawler cannot reach the page, it cannot become source [1]
Perplexity fetches its sources live

Its crawler exists to link you, not to train on you

Perplexity documents two crawlers. PerplexityBot is described as designed to surface and link websites in search results on Perplexity, and the same page states it is not used to crawl content for AI foundation models. Perplexity-User is the one that opens a specific page when a user asks for it.

If you dislike the idea of your writing training somebody's model but you would still like the traffic, this is the cleanest case on the web. You can turn away the training crawlers, GPTBot, ClaudeBot, CCBot, and still let PerplexityBot in without contradicting yourself. One of those choices costs you readers. The other does not.

It reads you live, so today's state is the one that counts

Because the fetch happens at question time, there is no cached copy from last month to fall back on. A page that is down, painfully slow, or built so that the text only appears after JavaScript runs can be skipped in that moment, and the answer gets written without you.

The practical version of that: whatever you want quoted should be sitting in the HTML your server sends, not assembled in the browser afterwards. You can check in one step. Open your page, view source, and search for a sentence you care about. If it is not in there, a crawler may never see it either.

Recency counts more here than elsewhere

Perplexity is built to feel current, and it shows in what it cites. On fast-moving topics a newer page can push out an older one with more authority behind it, where ChatGPT and Google's AI tend to stay with settled sources for longer.

You do not need to publish constantly to benefit. Update the pages that carry your best answers, and give them a real date the crawler can read, not a footer that says the current year because your template prints it automatically. A page with no honest date looks the same to a machine whether you wrote it last week or in 2019.

Write the passage it can drop into an answer

Perplexity builds an answer out of short extracts. The pages that get pulled in are the ones where somebody asked a plain question in a heading and answered it in the next two or three sentences, completely, without needing the paragraph before it for context.

That is also why thin pages rarely make it. A heading with one vague sentence under it gives the engine nothing to lift, and a 3,000 word essay with the answer buried in the middle gives it too much to dig through. Clear question, immediate answer, then the detail for the humans who keep reading.

Check that you are reachable

Start with your robots.txt, the text file at the root of your domain that tells crawlers what they may read. Open your domain followed by /robots.txt and look for PerplexityBot with a Disallow line under it, and for a User-agent: * group that disallows everything. Either one takes you out of the running.

Worth knowing: robots.txt is not the only gate any more. Some hosts and CDNs, Cloudflare among them, now offer a single switch that blocks AI crawlers at the network level before they ever reach your file. If your robots.txt looks clean and you still see no AI traffic, that setting in your hosting panel is the next place to look.

HowFindable scans your site for free and reports which AI crawlers your robots.txt turns away, along with the structure, freshness and schema signals the answer engines read. If you want the fixes, we generate the corrected files ready to upload.

Could Perplexity link you today?

Paste your URL and get two scores, one for Google and one for AI, plus the exact signals you are missing. If you want, we generate the fixed files for you, ready to upload. The scan is free.

Scan your site