How to rank in Perplexity: a 2026 guide for businesses

Perplexity is an answer engine. It searches the web in real time for each question, has a language model write the answer, and attaches numbered citations that link to the original sources.¹ For a business the citations are what matter: they carry the links, and they show the reader which sites the answer relied on. This guide covers what Perplexity offers in November 2026, how it finds and chooses sources, what makes a page more likely to be cited, and how to track the URLs it cites.

Crawler settings, IP verification and the 2025 dispute with Cloudflare are covered in Perplexity's crawlers explained, so this post only touches on them. How Perplexity compares with ChatGPT, Google's AI features and the other assistants is in the AI search guide.

Search is still the core product. It now sits alongside a research mode, an agent, a browser and a shopping service, and each of them reads the web, so each is a route by which your pages can reach a user.

Perplexity's help page for its PayPal offer named Wayfair and Abercrombie & Fitch among the Instant Buy merchants, along with thousands of others on BigCommerce, Shopware and Wix.⁸ When it launched its shopping assistant in November 2024, Perplexity also opened a free Merchant Program for large retailers to share product specifications, said joining could make a product more likely to be recommended, and said its shopping product cards were not sponsored.⁹ I could not find a current Perplexity page setting out the programme's terms, so treat that description as history. The groundwork that serves every AI shopping surface is in product data for AI shopping.

The publisher programme Perplexity launched in August 2025 is Comet Plus, a $5 subscription that is also included in Pro and Max. Perplexity says it passes the revenue to participating publishers, keeping a small share for compute, and allocates it by human visits, search citations and agent actions. Publishers apply through publishers@perplexity.ai.¹⁰ It matters to news and magazine publishers. For most other businesses it is not worth chasing.

There is no paid route into the answers. Perplexity ran a sponsored questions ad product in 2024,⁹ then phased advertising out, and in February 2026 decided to stop pursuing advertising revenue, ITdaily reported, citing the Financial Times.¹¹ Getting cited is organic work.

Perplexity answers from its own index. It says its Search API gives access to the same global-scale infrastructure that powers its public answer engine.¹² Its research team describes an index that tracks over 200 billion unique URLs, crawled by PerplexityBot within the limits sites set in robots.txt, on search infrastructure that processes 200 million queries a day.¹³ The index takes tens of thousands of update requests every second.¹²

Not every URL gets in, and not every page is refreshed at the same rate. A machine learning model predicts whether each candidate URL needs to be indexed, and models and heuristics decide which documents to keep hot, with documents from authoritative domains named as one priority. Refreshing pages already in the index competes with indexing new ones for the same budget.¹³ My reading is that a small site with few links will be revisited less often than a large one, as with any search engine.

There is a second, live route. Perplexity-User may visit a page to help answer a user's question and link to that page in the response. It generally ignores robots.txt, because a user requested the fetch.¹⁴

Perplexity also licenses some data. In March 2024 it began showing Yelp's maps, reviews and photos for local businesses, mainly restaurants and cafes, through Yelp's Fusion data programme.¹⁵ In January 2025 its hotel results began to carry Tripadvisor summaries, ratings and images.¹⁶ Paid plans add premium sources, which Perplexity describes as top financial data and trusted medical sources.² I have not found a more recent statement on the Yelp or Tripadvisor arrangements, so look at how answers in your own category are sourced before relying on either.

Perplexity's research team set out how its search system ranks results in September 2025.¹³ The stages below follow that account.

• Parsing. A content understanding module extracts the meaningful content of each page with rules that vary by site structure, aims to preserve the original structure and layout, and splits each document into self-contained spans that can be retrieved and ranked on their own.¹³
• Retrieval. The query runs against the index lexically, matching words, and semantically, matching meaning, and the two sets of results are merged into one list of candidates.¹³
• Prefiltering. Basic heuristics remove clearly non-responsive or stale content.¹³
• Ranking. Fast lexical and embedding-based scorers narrow the list, then more powerful cross-encoder rerankers produce the final order, scoring both whole documents and the spans inside them.¹³

Two practical points follow. The words on the page still count, because retrieval is partly lexical, so use the terms your customers use for the service, product or place. And each section is judged on its own, so a passage that answers a question without leaning on the paragraphs around it can be retrieved for that question even when the rest of the page covers something else.

Perplexity also marks some cited domains with a source label: Government, Academic or Trusted. Trusted covers sites that appear often enough in results to receive a label and that publish within their own area of expertise. Labels come from a source review process that rates whole websites and asks questions such as whether a site corrects its mistakes, says who wrote each piece, and keeps news separate from advertising and opinion. Perplexity says partnerships and payments do not affect a site's label, and that a label is not an endorsement.¹⁷ It does not say whether labels affect ranking. The review questions are worth meeting anyway, because a careful reader asks the same things.

Ordinary search rankings carry over to a degree. In a study published in August 2025, Ahrefs ran 15,000 long-tail queries through Google, Bing, ChatGPT, Gemini, Copilot and Perplexity. It found that 28.6% of the URLs Perplexity cited ranked in Google's top 10 for the same query, the highest share of the four assistants, against an average of about 12%.¹⁸ Ranking well in search helps, and most Perplexity citations still came from pages outside Google's top 10.

Allowing PerplexityBot and Perplexity-User

Perplexity recommends allowing PerplexityBot in robots.txt and permitting its published IP ranges.¹⁴ If you disallow a page, PerplexityBot will not index its text, though Perplexity may still index the domain, the headline and a brief factual summary.¹⁹ The usual failure is a CDN or firewall rule that blocks AI crawlers whatever robots.txt says, which when your CDN blocks AI crawlers covers. The robots.txt groups and IP checks for both Perplexity agents are in the crawlers post.

Server-rendered content

Vercel and MERJ's log analysis, published in December 2024, found that none of the major AI crawlers it measured rendered JavaScript, PerplexityBot included.²⁰ I have not seen Perplexity document JavaScript rendering since. Put prices, opening hours, service descriptions and anything else you want quoted in the HTML the server sends. Checking what crawlers see on a JavaScript site shows how to test it.

Pages that answer the question directly

Because Perplexity ranks self-contained spans, write sections that each answer one question and name the subject inside the section, so the passage still makes sense when lifted out. A heading such as "Boiler servicing prices in Leeds" followed by a sentence that gives the price range does more work than a paragraph that builds up to the number.

The closest thing to a controlled test on Perplexity is academic. The GEO paper (Aggarwal and others, KDD 2024) ran 200 queries through Perplexity with the source text uploaded as files. Adding quotations improved the paper's position-adjusted word count measure by 22% over the baseline, and keyword stuffing did 10% worse than the baseline.²¹ Those were lab conditions on an earlier version of the product, so I read them as a direction: specific, quotable, sourced writing does better than repetition.

Freshness and dates

Perplexity's pipeline drops stale content before ranking, and its index balances refreshing known pages against finding new ones.¹³ Its Search API returns a publication date and a last-updated date for each result,²² so dates are part of what the index holds about a page. Show both plainly on anything that dates, such as prices, guides and comparisons, and change the updated date only when the content changes.

Ahrefs' study of 17 million citations, published in July 2025, found that the pages Perplexity cited were on average about 250 days younger than those in Google's organic results, though still more than three years old.²³ Fresh content has an edge, and long-lived pages still get cited.

Third-party coverage and mentions

Much of what Perplexity cites is not on brand sites. Across more than 3.1 million US queries tracked in Ahrefs Brand Radar for November 2026, Reddit took 24.6% of the citations going to Perplexity's 50 most-cited domains, YouTube 14.2% and Wikipedia 7.2%. Amazon, Facebook, Healthline and The New York Times were also in the top ten.²⁴

For most businesses that means being discussed where Perplexity looks: useful answers in the Reddit communities your customers use, videos that show the product or service, reviews on the platforms that matter in your category, and coverage in trade and local press. For a restaurant or hotel, the Yelp and Tripadvisor data described above is part of the picture. I would only do this openly, as the business, because a thread exposed as planted does more harm than no thread. Brand signals in AI search goes further into this.

I have not found a Perplexity report for site owners that counts citations or clicks, so tracking is assembled from several places. The method for building a prompt set, sampling it and reading the results is in how to measure visibility in AI search.

Keep the traffic in proportion. SE Ranking's analysis of 101,574 websites with Google Analytics, covering January 2025 to April 2026, put Perplexity at 7.23% of AI referral traffic in 2026, third behind ChatGPT at 74.78% and Gemini at 11.56%, and all AI platforms together at 0.32% of website traffic.²⁸ Judge Perplexity on citations and on what its visitors do, as well as on volume.

Perplexity's record on crawling is argued over. In August 2025 Cloudflare alleged that Perplexity was changing its user agent and source networks to hide its crawling,²⁹ and Perplexity replied that Cloudflare had confused it with 3 to 6 million daily requests from BrowserBase, a third-party cloud browser service.³⁰ Both accounts are set out in the crawlers post.

In October 2025 Reddit sued Perplexity, SerpApi, Oxylabs and AWMProxy, alleging that they scraped Reddit content from search results and used it without a licence. Reddit said it created a post that only Google's crawler could reach and that Perplexity's answer engine soon reproduced its content. Perplexity said it would fight for users' rights.³¹ I have not seen a ruling reported, so I would not draw conclusions about Perplexity's sources from the case.

Comet's agent has its own legal fight. In March 2026 a federal court in San Francisco ordered Perplexity to stop using Comet's agent to make purchases for users on Amazon.³² In August 2026 Engadget reported that the Ninth Circuit had overturned that injunction, reasoning that the Comet user was the one accessing Amazon's computers, and the lawsuit continues.³³ For a retailer, the open question is whether an agent acting for a signed-in customer should be treated as that customer. The courts have not settled it.

• Check access first. Confirm that PerplexityBot gets normal responses from its published IP ranges, at the CDN as well as in robots.txt.
• Fix the pages that should be cited. For each service or product, one page whose sections answer the questions buyers ask, in server-rendered HTML, with visible dates.
• Earn mentions where Perplexity looks. Reddit, YouTube, review platforms and trade press carry a lot of weight in its citations.
• Track monthly. A small prompt set sampled by hand or through the API, Perplexity-User requests in your logs, and perplexity.ai sessions in analytics.
• Skip anyone selling guaranteed Perplexity placement. Perplexity has no ads to buy in its answers, and nobody outside it controls the ranking.
• Skip Comet Plus unless you are a publisher.