How Perplexity Chooses Its Sources, and How to Become One
Perplexity searches its own web index, ranks pages and short passages by exact words and meaning, then cites the best ones. To become a source, allow PerplexityBot and write clear, specific passages.
Perplexity chooses its sources by searching its own web index, ranking whole pages and the short passages inside them for how well they match the question, and then writing an answer that cites the passages it used. Its engineers say pages from authoritative domains get priority, and that both the exact words and the meaning of a passage count. To become a source, let PerplexityBot crawl your site, and write pages made of clear, specific, self-contained answers to the questions your buyers ask.
Perplexity is smaller than ChatGPT or Google, but it's one of the more transparent AI answer engines about how its search works. That makes it a useful place to learn what all of them reward. If the terms are new to you, start with generative engine optimisation.
Where does Perplexity get its information?
Perplexity answers mainly from its own search index, which its crawler, PerplexityBot, builds by visiting websites.
In a September 2025 engineering article about its Search API, Perplexity said its index "tracks over 200 billion unique URLs" and is updated constantly. Perplexity's crawler documentation says PerplexityBot exists to "surface and link websites in search results" and is not used to collect content for training AI models.
There's a second visitor too. Perplexity-User fetches a page when someone's question needs it right now, for example when they paste in your link or ask about something very recent. Perplexity says this fetcher "generally ignores robots.txt rules", because a person asked for the page.
So there are two ways into a Perplexity answer: be in the index already, or be fetched live. The index is the one you can plan for.
How does Perplexity rank and pick sources?
Perplexity ranks individual passages, not just pages, using a mix of keyword matching, meaning-based matching and a final AI reranking step.
Its engineering article describes the pipeline in some detail. Here is what it says and what each step means for a business website.
| What Perplexity says it does | What it means for your pages |
|---|---|
| Uses a model to predict whether a URL is worth indexing at all | Thin, duplicate or empty pages may never make it in |
| Prioritises "documents from authoritative domains" | Mentions and links from credible sites help you get indexed and kept |
| Splits documents into "self-contained spans" ranked on their own | Each section must make sense if lifted out alone |
| Retrieves by both lexical (keyword) and embedding (meaning) matching | Use the exact words buyers use, and say clearly what you mean |
| Uses "cross-encoder reranker models" for the final cut | The passage that answers the question most directly wins |
| Aims to capture content "high quality in both substance and form" with its structure preserved | Clean headings, lists and tables survive and help |
| Keeps the index "constantly updated" | Current, dated information has an edge for recent questions |
The model then writes an answer from the top passages, with numbered citations next to the claims that link to the sources. People can click any citation to check it, which is why being cited can send real visitors.
What Perplexity doesn't tell us
Perplexity doesn't publish ranking weights, a list of favoured websites, or how its answer model chooses between equally good passages. Plenty of blogs claim to know, often with confident percentages. Treat those as guesses. The safe reading of Perplexity's own description is simple: be indexable, be credible, and be the clearest answer to a specific question.
Why passages matter more than pages
Because Perplexity scores passages on their own, a page wins citations one section at a time.
Imagine two solar companies in Kumasi. The first has a long "About our services" page that mentions inverter sizes somewhere in the fourth paragraph. The second has a section headed "What size inverter does a three-bedroom house in Kumasi need?" that opens with a direct answer, gives the typical range, explains what changes it, and lists the appliances that push it up.
Ask Perplexity that question and the second company's section is the obvious span to cite. It matches the words, matches the meaning, and stands on its own. The first company's page might be indexed, but nothing in it is quotable.
This is the same writing approach that works for Google's AI Overviews and ChatGPT. We cover it in full in how to write content that AI answers quote.
How to become a Perplexity source: step by step
You can't make Perplexity cite you, but you can remove every reason it wouldn't. These steps are in order of priority.
- Allow PerplexityBot. Check yoursite.com/robots.txt for any rule that blocks
PerplexityBotor all bots. Perplexity says changes can take up to 24 hours to apply. - Check your firewall and CDN. Perplexity's documentation recommends allowing its published IP ranges and matching both the user agent and the IP address in firewall rules. If your host or CDN has an "AI bot blocking" switch, make sure it isn't stopping PerplexityBot.
- Make sure your important pages exist and are linked. Every service, location and product line needs its own page, linked from your menu or other pages, so a crawler can find it.
- Put the facts in plain text. Prices, service areas, delivery times and opening hours inside images, Instagram posts or PDFs are harder to retrieve and quote.
- Write question-led sections. Use the questions buyers actually ask as headings, and answer each one in the first sentence or two below it.
- Use the buyer's words. Lexical matching rewards exact terms. If customers say "MoMo", "pay on delivery" or "Spintex", use those words, not just "mobile payments" or "east Accra".
- Be specific. Numbers, names, places and dates make a passage more useful than a vague one. "Delivery to Tema within 24 hours, GH₵ 30" beats "fast, affordable delivery".
- Keep pages current. Show when a page was last updated and change it when facts change. Old prices are worse than no prices.
- Build authority off your site. Get listed in credible directories, earn press coverage, and appear on the comparison and "best of" pages Perplexity already cites. See why AI recommends brands from "best of" lists and how to earn backlinks in Ghana.
- Add structured data that matches the page. Organisation, FAQ and breadcrumb markup help machines read your facts. Perplexity hasn't said it uses schema, so treat it as good practice rather than a Perplexity lever. See schema markup for small businesses.
Find out what Perplexity cites today
The fastest way to learn what Perplexity wants in your industry is to ask it your buyers' questions and study the sources it shows.
Spend 30 minutes on this:
- Write down ten questions a buyer asks before choosing a business like yours, such as "best event decorator in Accra" or "how much does a borehole cost in Ghana".
- Ask each one in Perplexity, from Ghana if your customers are in Ghana.
- Click through the cited sources. Note the type of page (directory, news article, blog, competitor's service page, Reddit thread) and what the cited section looks like.
- Ask whether your page could have been the source. If not, why not: missing page, no direct answer, no specifics, no credibility?
- Repeat monthly and record the results. See how to measure your visibility in ChatGPT, Gemini and Perplexity.
For many Ghanaian categories, the cited sources are thin or foreign, which is an opening for a local business that publishes clear answers. See AI search optimisation in Ghana.
You'll usually find one of three gaps. You have no page for the question. You have a page, but it doesn't answer the question directly. Or the cited sources are third-party lists you're not on. Each gap has a clear fix.
What doesn't work
Some tactics waste money, or risk your reputation.
- Keyword stuffing. Rerankers judge whether a passage answers the question, not how often a word repeats.
- Mass-produced AI pages. Perplexity uses a model to decide which URLs are worth indexing at all. Pages that repeat what's already out there give it little reason to keep them.
- Fake reviews and planted forum posts. They're often easy to spot, and they damage trust with the people who read them.
- Blocking crawlers to "protect" content, then expecting citations. If PerplexityBot can't read the page, it can't index it.
Common questions
Does Perplexity use Google's results?
Perplexity describes its own index and crawler as the base of its search. It doesn't publish a full list of every data source it uses, so we don't assume more than that.
Does Perplexity train AI models on my website?
Perplexity's documentation says PerplexityBot "is not used to crawl content for AI foundation models", and that Perplexity-User isn't used for training either.
Can I block Perplexity-User?
Perplexity says it generally ignores robots.txt because a person requested the page. If you need to block it, you'd have to do so at your firewall, but for most businesses those visits are potential customers.
How long until Perplexity cites my new page?
There's no fixed timeline. Robots.txt changes apply within about a day, but crawling, indexing and being chosen over other sources can take weeks or longer, and isn't guaranteed.
Does llms.txt help with Perplexity?
Perplexity hasn't said it uses llms.txt. It's cheap to add, but crawlable pages with clear answers matter far more. See what llms.txt is.
Is Perplexity worth the effort for a Ghanaian business?
The work that earns Perplexity citations also helps you in ChatGPT, Gemini and Google, so it's rarely wasted. See where AI assistants get their information and how to get recommended by ChatGPT.
If you want help becoming the answer Perplexity and other AI assistants cite, see our AI search optimisation service or book a free strategy call.