- Perplexity says its index tracks over 200 billion unique URLs and splits documents into "self-contained spans" that are retrieved and ranked one by one.
- It returned 19.08 links per answer in Slate's 1,000-question test, against 12.49 for Claude and 3.10 for ChatGPT.
- Google rank carries over strongly: Perplexity cited 86.9% of pages ranking first on Google for a question and 46.0% of pages ranking ninth.
- Reddit was 21.6% and YouTube 20.8% of citations among Perplexity's top 50 sources in Ahrefs' September 2026 data.
- PerplexityBot builds the search index and isn't used to train foundation models, so blocking it buys you no training opt-out, only invisibility.
What does it mean to rank in Perplexity?
To rank in Perplexity is to be cited as a source in its answers, ideally early in the list, and for commercial questions, to be one of the brands the answer recommends. Perplexity shows its sources with every answer, so unlike ChatGPT you can see exactly who won and who didn't.
The volume is real. Perplexity's CEO said it handled 780 million queries in May 2025, about 30 million a day9. And because it shows so many links, a niche page has a better shot here than in any other assistant.
This guide is about Perplexity specifically. For the comparison across ChatGPT, Gemini, Claude and Copilot, read how AI engines choose their sources.
How does Perplexity choose which sources to cite?
Perplexity cites from its own index, which it crawls with PerplexityBot and ranks with a multi-stage pipeline. In September 2025 it opened that index to developers and published an engineering write-up of how it works3. That's the best primary source on Perplexity ranking, and most guides ignore it.
What gets into the index
The index "tracks over 200 billion unique URLs"3. A machine learning model decides whether a URL is worth indexing and when to recrawl it, calibrated to "both the importance and likely update frequency of the specific URL." Perplexity keeps some documents "hot," prioritizing "documents from authoritative domains, documents involving undercovered topics"3. The launch post adds that its systems process "tens of thousands of index update requests" every second2.
Two practical consequences. Pages that change often and matter get recrawled more, so a page that never changes may sit stale in the index. And there's no URL submission tool: Perplexity offers no webmaster console, so crawl access is the only way in.
How pages are split and scored
Perplexity's parser decomposes documents "into self-contained spans, each of which can be individually retrieved and ranked"3. Retrieval is hybrid: it queries the index both lexically and semantically and merges the results. Prefilters "remove clearly non-responsive or stale content." Early stages use lexical and embedding-based scorers, then "more powerful cross-encoder reranker models" score candidates at both the document and sub-document level3.
Read that as a writing brief. Your page competes paragraph by paragraph, a section that only makes sense after the previous one is a weak candidate, and stale pages can be filtered out before any model reads them. This is the chunking problem made concrete.
User behavior feeds the rankers
Perplexity says its stack learns from "the rich signal produced by the millions of user requests we serve each hour," combining AI evaluations "with live human feedback" to train its embedding and ranking models3. It doesn't say which signals. Don't build a strategy on guesses about clicks; do note that pages people find useless are unlikely to keep their place.
The two crawlers
Perplexity's crawler documentation lists two agents1:
| Agent | What it does | robots.txt |
|---|---|---|
| PerplexityBot | Surfaces and links websites in Perplexity search results. "Not used to crawl content for AI foundation models." | Respected. Changes can take up to 24 hours to apply. |
| Perplexity-User | Fetches a page when a user's question needs it | "Generally ignores robots.txt rules," because a user requested the fetch |
Perplexity's help center adds a detail worth knowing: if a page is blocked, "we may still index the domain, headline, and a brief factual summary," and it "partners with third-party crawlers to help build our search index"4. A robots.txt block makes you a weak source, not an absent one.
Perplexity-User's behavior has been disputed. In August 2025 Cloudflare said Perplexity used undeclared crawlers to get around sites' no-crawl rules and removed it from its verified bot list8. Perplexity replied that Cloudflare had confused its traffic with unrelated requests from a third-party browser service. Either way, if you run Cloudflare, check how its bot rules treat Perplexity before assuming you're reachable.
What does the data say Perplexity cites?
Two large datasets give a consistent picture. Slate's study ran 1,000 B2B software questions through each engine with web search on, captured 3 August 20265. Perplexity linked a source on every question, 19.08 links per answer on average. It cited 86.9% of pages ranking first on Google for the question and 46.0% of pages ranking ninth. Google rank had an effect size of 0.307, while every on-page attribute Slate measured stayed under 0.10. In Perplexity, where you rank matters more than how long or polished the page is.
By page format, Perplexity cited ranking pages at these rates5:
| Page type | Perplexity cite rate |
|---|---|
| Pricing | 77.1% |
| Comparison | 73.8% |
| Listicle | 71.7% |
| General | 69.5% |
| Definition | 68.1% |
| How-to | 66.6% |
Community content is the other half. Slate counted 1,051 Reddit citations across 617 of its 1,000 questions, and Perplexity cited 74.2% of ranking Reddit, Quora and Stack Overflow pages5. Ahrefs' Brand Radar data for September 2026, covering more than 3.1 million US queries, put Reddit at 21.6% and YouTube at 20.8% of the citations going to Perplexity's 50 most-cited sources, ahead of Wikipedia at 6.3%6. Ahrefs called Reddit "the most concentrated top source of any assistant we looked at."
The 8-step plan to rank in Perplexity
Steps 1 to 3 are a week of setup. Steps 4 to 7 repeat. Each ends with a test.
1. Write 30 prompts, half of them "best" and "vs" questions
Perplexity users ask research questions, so mix definitions and how-tos with the commercial prompts that decide deals: category ("best SOC 2 automation tool for a 50-person startup"), comparison ("Vanta vs Drata"), alternatives and pricing. Pricing and comparison pages had the highest cite rates in the Slate data, so include prompts that reward them.
Done when: you have 30 or more prompts, tagged by type, with at least ten commercial ones.
2. Run a baseline and export the sources
Run each prompt three times, signed out or in a fresh account, and copy every source from the Sources panel in order. With about 19 links per answer, this gives you a rich list fast. Count a citation only if it appears in two of three runs.
Done when: 90 runs are logged with cited URLs in order, your citation rate per prompt type, and the domains that appear most.
3. Let PerplexityBot in, at robots.txt and at the firewall
Allow PerplexityBot in robots.txt (our robots.txt AI checker shows what your file says). Then check the firewall, where most blocks live. Perplexity's docs tell Cloudflare users to create a WAF custom rule matching the PerplexityBot or Perplexity-User user agent and the IPs in perplexitybot.json and perplexity-user.json, with the action set to Allow1. Grep 30 days of logs for PerplexityBot and match the IPs.
Done when: verified PerplexityBot requests got 200 responses on your money pages within the last 30 days.
4. Check what Perplexity's index returns for your queries
Perplexity's Search API exposes the same index its answers draw on, and each result comes back with a title, URL, the extracted snippet, a publication date and a last-updated date7. Query your core phrases and see whether your page comes back, which passage Perplexity extracted, and what date it thinks the page has. A wrong or missing date on a page you updated last week tells you the index hasn't caught up.
Done when: you know, for each money query, whether your page is in the top 10 results and which passage was extracted.
5. Rewrite sections as self-contained spans
Because Perplexity ranks spans, make each H2 section stand alone. Put the question in the heading, the answer in the first sentence, and the number with its source in the next. Replace "as mentioned above" and "this approach" with the noun. Keep tables in HTML, since the parser is graded on preserving "the original content structure and layout"3. The citability grader flags sections that lean on context.
Done when: every H2 on your money pages answers its heading in the first sentence and can be quoted without the paragraph before it.
6. Keep money pages genuinely current
Perplexity's prefilters drop stale content, and recrawl frequency follows how often a page changes3. Its API tracks publish date and last-modified date as separate fields7. So update pricing, comparison and "best" pages when facts change, show a real updated date, and don't fake freshness by changing the date alone.
Done when: each money page has an owner and a review cycle, and its visible updated date matches a real change.
7. Show up where Perplexity's community sources live
From your baseline, list the Reddit threads and YouTube videos Perplexity cites for your prompts. Answer the threads where you have something specific to add, with your affiliation stated. For video-heavy prompts, a short walkthrough with a title that matches the question is a realistic way into a source list that YouTube already dominates.
Done when: you've contributed to the five most-cited threads for your prompts, and each video-heavy prompt has a matching video or a reason not to make one.
8. Re-run monthly and compare sources
Same prompts, same method, fixed day. When your citation rate moves, check which URLs entered or left the source lists before you credit or blame your own changes.
Done when: the re-run is scheduled and every report shows citation rate by prompt type with run counts, next to last month.
Worked example: a compliance software company
This is a hypothetical company, "Attestly," which sells SOC 2 automation to startups. We haven't run these prompts; the example shows how the checks and decisions fit together, not results.
The index check. Attestly queries the Search API with one of its money phrases7:
curl -X POST 'https://api.perplexity.ai/search' \
-H "Authorization: Bearer $PERPLEXITY_API_KEY" \
-H 'Content-Type: application/json' \
-d '{"query": "SOC 2 automation pricing for startups", "max_results": 10}'It reads each result's url, snippet and last_updated.
The before and after. Attestly's pricing page opened like this:
Before: Pricing that scales with you. Every company is different, which is why
we tailor our plans. Talk to our team to learn more about what's included.Nothing in that paragraph can answer "how much does SOC 2 automation cost." The rewrite answers the heading in one quotable span:
After: Attestly costs $X a year for companies under 50 employees and $Y for
50 to 200, billed annually. Both plans include SOC 2 Type I and Type II,
policy templates and auditor access. ISO 27001 is a $Z add-on.(The prices are placeholders. Use your real ones; a span with no number is the problem being fixed.)
Reading the baseline.
- If "best SOC 2 automation tools" cites three listicles and two Reddit threads that skip Attestly, the work is getting onto those pages and into those threads. Its own homepage isn't competing for that prompt.
- If Attestly ranks top three on Google for a query but Perplexity never cites it, it checks the firewall logs first. A Cloudflare bot rule challenging PerplexityBot is the usual cause.
- If the Search API returns Attestly's page with an extracted snippet from the cookie banner or navigation, the main content isn't in clean HTML. Fix the template, not the copy.
- If the cited pages for "Vanta vs Drata" never mention Attestly, a "Vanta vs Drata vs Attestly" comparison page with a real table is the move, since comparison pages had a 73.8% cite rate.
What doesn't get you cited by Perplexity
Blocking PerplexityBot to stop AI training
PerplexityBot isn't a training crawler1. Blocking it removes you from search results while doing nothing for training. If your worry is training, deal with the crawlers that do train; our AI crawler list covers which is which.
Chasing the "59 ranking factors"
In 2025, an independent researcher published what he described as Perplexity's ranking patterns, inferred from browser-level traffic. Search Engine Land covered it and called it "unverified at this point"10. Some of it, such as authoritative-domain priority and freshness, matches Perplexity's own write-up. Much of the rest, like topic boosts and early-click windows, may describe feed features rather than answer citations. Build on what Perplexity has published.
Buying ads
Perplexity tested ads as sponsored follow-up questions in November 2024 and wrote that it would avoid "duplicating the SEO industry where people are implementing arbitrary tactics to improve their rankings"11. In February 2026 the Financial Times reported that Perplexity was phasing ads out over trust concerns, as TechCentral relayed12. Ads never bought a spot in the source list either way.
Changing the date without changing the page
Freshness matters, but Perplexity tracks last-modified separately and its parser reads the content. A new date on an unchanged pricing page that still lists last year's plans is a page a buyer will catch, and a stale-content filter is designed to catch.
Products and publishers: the programs that exist
If you sell physical products, Perplexity's shopping launch said product cards "aren't sponsored," and its Merchant Program, free for merchants, increases the "chances of being a 'recommended product'" because products are then in its index13. In November 2025 it made shopping free for all US users with PayPal checkout, with retailers staying merchant of record14. Large retailers should apply; everyone else should make sure their product pages are crawlable with prices in the HTML.
Publishers have a separate route. The Publishers' Program, launched in July 2024, shares revenue when Perplexity earns money from an interaction that references a publisher's content15. Revenue sharing doesn't change ranking, and Perplexity hasn't said it does.
How do you measure your Perplexity rankings?
Track citation rate (share of runs citing your domain), average source position when cited, brand mention rate on commercial prompts, and the top cited domains, all with run counts. Because Perplexity always shows sources, this is the easiest engine to measure by hand, and the most practical one to automate. Our comparison of Perplexity rank trackers covers what each tool measures and costs, and the guide to measuring AI visibility covers sample sizes.
For traffic, look for sessions from perplexity.ai in GA4. When we checked, perplexity.ai served a strict-origin referrer policy, so clicks arrive with the domain as referrer and no path; you'll see which of your pages got the visit, not which answer sent it. For a free first read across engines, the Arobis AI visibility checker shows whether AI engines name your brand for your category's questions.
How long does it take to rank in Perplexity?
Technical fixes can show up quickly, because the index updates continuously and Google rank carries over strongly. Robots.txt changes reach PerplexityBot within about 24 hours1. Page rewrites show up once the page is recrawled, which depends on how important and changeable Perplexity's model thinks it is. Getting into cited Reddit threads can count within days; getting onto other people's listicles takes their editorial cycle. Give it a month for technical and on-page work, a quarter for off-site.
The work overlaps with other engines. Claude shared more cited pages with Perplexity than any other engine pair in the Slate test5, so see how to rank in Claude, plus our guides for ChatGPT, Gemini and Google AI Overviews.
Frequently asked questions
How does Perplexity decide which sources to show?
It searches its own index of over 200 billion URLs, mixing keyword and semantic retrieval, filters out stale or off-topic content, then ranks candidates with cross-encoder models at both page and passage level. Perplexity says user feedback also trains its rankers. In practice, pages that rank well on Google and answer the question in a self-contained passage get cited most.
Does Perplexity use Google results?
Perplexity says it uses its own index, built by PerplexityBot and partly by third-party crawlers. Still, Google rank and Perplexity citations line up closely: in Slate's 2026 test, Perplexity cited 86.9% of pages ranking first on Google for the question and 46% of pages ranking ninth. Ranking well on Google is a strong predictor, even if it isn't the source.
Should I block PerplexityBot?
Only if you don't want to appear in Perplexity at all. PerplexityBot indexes pages for search and isn't used to train foundation models, so blocking it doesn't protect your content from training. Perplexity-User, which fetches pages on a user's request, generally ignores robots.txt. Even blocked pages may keep their domain, headline and a short summary in the index.
Does Reddit help you rank in Perplexity?
Yes, more than in any other assistant. Reddit took 21.6% of citations among Perplexity's top 50 sources in Ahrefs' September 2026 data, and Slate found Perplexity cited 74.2% of ranking Reddit, Quora and Stack Overflow pages. Contributing useful, disclosed answers to the threads Perplexity already cites for your prompts is a legitimate way into its source lists.
How do I submit my site to Perplexity?
You can't, directly. Perplexity has no webmaster console or URL submission tool. Allow PerplexityBot in robots.txt and at your firewall, using Perplexity's published IP ranges, and its crawler schedules pages based on how important they are and how often they change. Retailers can apply to the free Merchant Program, and publishers can contact the Publishers' Program.
How long does it take to show up in Perplexity?
Robots.txt changes reach PerplexityBot within about 24 hours, per its docs, and Perplexity processes tens of thousands of index updates per second. A page that's crawlable and already ranks on Google can be cited within weeks. Getting named on third-party roundups depends on their update cycles, so expect a quarter for off-site work to show.
Sources
- Perplexity. "Perplexity crawlers" documentation.
- Perplexity. "Introducing the Perplexity Search API." 25 September 2025.
- Perplexity Research. "Architecting and evaluating an AI-first search API." 25 September 2025.
- Perplexity Help Center. "How does Perplexity follow robots.txt?"
- Slate. "How to get cited by ChatGPT, Perplexity, Claude and Gemini" (1,000-question study, data captured 3 August 2026). 27 August 2026.
- Ahrefs. Most-cited domains in Perplexity (Brand Radar, September 2026 data).
- Perplexity. Search API quickstart.
- Cloudflare. "Perplexity is using stealth, undeclared crawlers to evade website no-crawl directives." 4 August 2025.
- TechCrunch. "Perplexity received 780 million queries last month, CEO says." 5 June 2025.
- Search Engine Land. "How Perplexity ranks content: research uncovers core ranking factors." 5 August 2025.
- Perplexity. "Why we're experimenting with advertising." 12 November 2024.
- TechCentral.ie. "Perplexity temporarily halts ads over concerns about user trust." 19 February 2026.
- Perplexity. "Shop like a Pro." 18 November 2024.
- Perplexity. "Shopping that puts you first." 25 November 2025.
- Perplexity. "Introducing the Perplexity Publishers' Program." 30 July 2024.
AI Ranked Editorial. "How to rank in Perplexity and get cited in its answers." AI Ranked, October 11, 2026. https://airanked.ai/guides/how-to-rank-in-perplexity