In short
Perplexity searches the web in real time and links numbered citations to its sources, but it does not publish how it ranks them. One small capture of its live data shows it retrieving 10 to 15 sources per query and citing only some, with the winners changing by query type. Claims about freshness and Reddit are widely repeated, but the figures are weaker than they look. The confirmed first step is simple: allow PerplexityBot in robots.txt and keep your content crawlable.
What Perplexity confirms
Very little, and that is worth saying plainly. Its help centre says it searches the internet in real time, gathers information from what it calls authoritative sources, compiles an answer, and includes numbered citations linking to the original pages. It does not explain how it scores or ranks those sources.
Its crawler documentation adds two useful facts:
- PerplexityBot surfaces and links websites in search results, and Perplexity says it is not used to crawl content for AI foundation models. It advises allowing this bot if you want to appear.
- Perplexity-User fetches pages for live user questions and “generally ignores robots.txt rules”, since a person requested the page.
Everything beyond that is inference from the outside, including most of what you will read in guides like this one.
Numbered citations
PerplexityBot and Perplexity-User
Reddit appears often
Citations per answer
How Perplexity picks sources, step by step
The closest thing to a look inside comes from Suganthan Mohanadasan, who recorded the data Perplexity streams to the browser while it answers, rather than studying the finished answers. It is a small sample: eight captures from one Pro account in Dubai in June 2026, with three rechecks in July. The author treats the field names and sequence as firm and the counts as direction only, and so should you.
The captures show this sequence:
- It classifies the query. A scorecard of 16 signals, each with a fixed threshold, decides what kind of query this is and whether maps, video or images appear. On the local query, the places signal scored 0.996 against a threshold of 0.85.
- It searches the web. All seven query types triggered a live search, even a basic how-to. Six ran a single search; the local query ran four.
- It gathers candidates. Each standard query retrieved 10 to 15 sources.
- It cites only some of them. Between 2 and 9 of those sources earned a citation on the queries where the count could be read.
- Deep Research reads further. It fetched a few pages in full, and one comparison page took 20 of the 30 citation markers.
The table is the useful part, because the kind of page that wins changes with the query.
| Query type | Example | Retrieved | Cited | What got cited |
|---|---|---|---|---|
| Commercial | best AI SEO tools 2026 | 10 | 6 | Recent listicles and mid-tier SEO blogs |
| Comparison | Ahrefs vs Semrush | 10 | 6 | ahrefs.com was the most-cited domain |
| News | latest Google algorithm update | 10 | 3 | Only Google's own properties |
| Local | coffee near DIFC | 15 | 5 | Business listings from maps, no editorial listicles |
| Shopping | earbuds under $150 | 10 | 2 | YouTube and a niche brand's review page |
| How-to | change a flat tyre | 15 | 9 | A mix that included YouTube videos |
So being retrieved is not the same as being cited, and neither a big brand nor Reddit is a sure thing. Semrush was retrieved and passed over on the commercial query, while three Reddit threads were retrieved and ignored on the shopping query. No ranking score appears in the data, so what sorts the candidates remains hidden.
What citation studies report
Researchers run many prompts and record what gets cited. Their findings are useful, as long as you do not treat them as Perplexity's actual algorithm.
Reddit is prominent, but the headline figure is easy to misread. Profound's citation data, covering August 2024 to June 2025, put Reddit at 6.6% of all Perplexity citations and 46.7% of the share among its ten most-cited sources. Profound says the second figure does not represent overall citation volume. You may see both credited to newer studies, but this is the earliest source we could find, so the data is older than it looks.
Freshness probably helps, but the evidence is thinner than it sounds. Searchless reports an 82% citation rate for relevant content from the last 30 days, attributed to Qwairy. We could not find Qwairy's original report or its method, so treat that as a claim, not a measurement. The captures above point both ways: recent listicles won the commercial query, yet a three-day-old Search Engine Land article lost the news query to Google's own pages. A reasonable reading is that freshness helps on time-sensitive queries but does not beat a primary source.
Citations per answer vary widely. The Searchless article gives 21.87 per answer, from a Qwairy analysis of 118,000 responses. Nightwatch reports 7.5 domains per answer in April 2026, down from 11.8 in November 2025. Citations and domains are different things, and methods differ, so these cannot simply be compared.
Published “ranking weights” are estimates. You may see precise percentages for relevance, authority and freshness. RankAI lists one such breakdown, attributed to Authority Tech rather than to Perplexity. Treat exact weights as somebody's model, not as a specification.
Why the numbers disagree
Citation studies measure a moving system, and differ in several ways:
- Prompt selection. Studies of commercial prompts are likely to find more review sites and forums than studies of technical ones.
- What counts. Total citations, unique domains and top sources all give different answers.
- Timing. Results shift as the product changes, and these studies age fast.
- Incentives. Many are published by companies selling AI visibility tools.
So use them for direction, not decimals. If a recommendation depends on one exact figure, be sceptical of it.
How to get cited by Perplexity
No method guarantees a citation. These steps hold up whichever study turns out closest, and most help with other AI engines and ordinary search as well.
Make sure Perplexity can read your pages
Allow PerplexityBot in robots.txt, and keep key content in the initial HTML. Perplexity says robots.txt changes can take up to 24 hours to apply. Check that your CDN or firewall is not blocking the bot either: its documentation publishes IP ranges for allow rules, and warns that Cloudflare's Bot Fight Mode can still challenge traffic that a custom rule would otherwise skip. Our AI crawler list covers the settings, and should you block AI crawlers? covers the trade-off.
Answer the question directly and early
Put a clear, self-contained answer directly below each question-style heading. A passage that makes sense when lifted out on its own is much easier to cite. See how to write content AI answer engines quote.
Update pages when the facts change
Refresh your important pages when facts change, and show a visible updated date. Changing a date without changing the content is not a real update, and it is the sort of shortcut that tends not to last.
Add details other pages lack
Original numbers, current prices, first-hand testing and named experts all give an engine a reason to cite you rather than the ten similar pages. This is usually more productive than any technical tweak.
Match the page to the query type
Look at what wins for your own queries, because the format varies. In the captures, a vendor's own page won its comparison query, map listings won the local one and video appeared on the shopping and how-to queries. So a local business may gain more from accurate map listings than from another blog post, and a product company may gain more from a useful comparison page.
Be present where it already looks
If Reddit, review sites, GitHub or trade press dominate the answers for your topic, a good presence there may do more than another article. Participate as a real person and follow each community's rules, because promotional spam tends to backfire.
Track the prompts that matter
Choose 20–30 questions your customers really ask. Run them on a schedule and record who gets cited, then study the pages that win. How to measure whether AI engines are citing you sets out a method. Run each prompt more than once, as answers vary.
Is Perplexity worth the effort?
Check before investing. Look at your analytics for Perplexity referrals using AI referral tracking, and test whether its answers actually influence your customers' decisions. For some technical and research-heavy audiences, it may matter a great deal; for many local businesses, Google is still far more important. The good news is that the work overlaps heavily, so you rarely have to choose. To compare engines, see how ChatGPT chooses which websites to cite and our Perplexity guide.
If you want to know how your brand currently appears in AI answers, see our GEO service or book a call.
Common questions
How do I get cited by Perplexity?
Allow PerplexityBot, keep content in the initial HTML, answer questions directly, keep important pages current, and include original detail. Check which page types win for your own queries, since that varies. Presence on sites Perplexity already cites for your topic can also help. No method guarantees a citation.
Does Perplexity cite every source it finds?
No. In one small capture of its live data, it retrieved 10 to 15 sources per query and cited between 2 and 9 of them where the count could be read. Large brands and Reddit threads were sometimes retrieved and not cited. The sample is too small to treat these counts as averages.
Does Perplexity use Google or Bing results?
Perplexity describes searching the web in real time but does not publish its full retrieval stack. Many guides recommend being indexed by both Google and Bing as a precaution, which is sensible regardless, but treat specific claims about its sources as unconfirmed.
Why does Perplexity cite Reddit so often?
Perplexity has not explained this. Profound's data for August 2024 to June 2025 put Reddit at 6.6% of all citations and 46.7% of the share among its ten most-cited sources, and that data is now dated. Forum threads often contain direct answers and first-hand experience, which may be part of the reason, but this is inference.
Does blocking PerplexityBot remove me from Perplexity?
It can reduce your visibility, as Perplexity says it should be allowed for sites to appear in search results. The separate Perplexity-User fetcher generally ignores robots.txt for live user requests, so blocking it completely needs firewall rules.
