This article explains how to rank in Perplexity in 2026, which for a B2B company means being recommended: your site cited and your company named. Three things decide it: whether Perplexity's two bots can reach your pages, whether the sources it cites name you, and whether your pages carry the buyer's wording in a passage that stands alone. Perplexity ranks pages, but a reader sees one answer with sources that change each time the question is asked, so there is no position to win. The measure is the share of a fixed set of buyer prompts that cite your site and name your company. In October 2026 Perplexity held 10.34% of US AI chatbot market share and 15.04% worldwide (Statcounter). The UK share was 4.0% in July 2026 (Statcounter), so a UK or EU company adds it after ChatGPT and Google's AI answers.
What ranking in Perplexity means
Ranking in Perplexity means your page is among the sources Perplexity retrieves for a buyer's question. Retrieved means the pages Perplexity reads while it writes the answer. Your page is then among the few sources Perplexity cites, the pages it links as numbered sources under the answer. And your company is named in the answer text.
Peec AI, the AI visibility tracking tool we use, finds that "64% of retrieved URLs are never cited in the answer at all" (Peec AI, retrieval vs citation).
Perplexity runs its own index, its own catalogue of web pages, "covering hundreds of billions of webpages" (Perplexity Search API announcement). It divides each document "into fine-grained units", passages in this article, and scores each one compared with the question.
Since September 2026 retrieval and ranking run on Photon, "a new in-house retrieval and ranking engine" (Perplexity, Photon). Photon is built on an inverted index, a lookup table from each word to the pages that contain it. So Perplexity does rank pages, even though no reader sees the ranking.
Perplexity, Peec AI, Agenzy.
Our GEO metrics guide, written on August 2026 data, told readers to add Perplexity once their own analytics showed buyers using it. The October 2026 share changes that advice.
For a B2B company the competition for these citations is still small. In our 2026 AI search visibility study, the median brand was named in 1.5% of Perplexity answers (n=85). On ChatGPT it was 6.5% and on Google AI Overviews 5.4%.
One Agenzy client, a B2B hardware manufacturer, went from 0% of Perplexity answers in May 2026 to 30% in September 2026 on its tracked prompt set. Over one seven-day window its own site was the source Perplexity opened most often in its category. That was 208 chats, compared with 170 for the next source (a GEO report example from that project).
The sections below cover how Perplexity finds and reads pages, crawler access, the sources it cites, freshness, the passage it quotes, and measurement.
How Perplexity finds and reads pages
A page enters Perplexity's answers by one of two routes. PerplexityBot crawls it into the index ahead of time, or Perplexity-User fetches it live when a person asks. Perplexity describes PerplexityBot as "designed to surface and link websites in search results on Perplexity" (Perplexity, bots).
Perplexity runs a web search for every question. The query it shows, meaning the words it searches for, is the question as the buyer typed it. It may refine queries internally without showing them.
In three client projects we track in Peec AI, over the 30 days ending 8 November 2026, Perplexity searched in every chat. One chat is one answer to one prompt.
Perplexity searched in all 600 chats for a B2B sales-outsourcing company selling in the US and Canada. It was all 9,300 for a consumer sports-equipment brand in four English-speaking markets, and all 2,000 for a B2B video-production company in Lithuania. On the same prompts ChatGPT searched in 91.5%, 85.1% and 83.0% of its chats, in the same order.
The sales-outsourcing project is the closest of the three to a B2B software vendor's buyers, and its queries show the pattern. For the same prompt on the same day, Perplexity searched "Average price of a B2B qualified meeting", one query in the buyer's words. ChatGPT ran three queries such as "2026 average cost per qualified B2B sales meeting appointment setting benchmark cost per meeting".
Peec AI data.
Crawler access and firewall blocks
Allow both Perplexity bots in your firewall and PerplexityBot in robots.txt, and check the CDN's security log. The CDN is the service, Cloudflare for example, that stands between your website and its visitors and can filter who gets in. The block that keeps a B2B site out of Perplexity usually sits there. Our list of AI crawlers has the exact user-agent strings.
Perplexity-User, the fetcher that acts on a user's request, "generally ignores robots.txt rules" (Perplexity, bots). Changes to robots.txt "may take up to 24 hours" to take effect. Perplexity also states that "Cloudflare's Bot Fight Mode can't be skipped with WAF custom rules" (the WAF is the web application firewall).
Cloudflare removed Perplexity from its verified bots on 4 August 2025, after reporting that Perplexity used undeclared crawlers when its declared ones were blocked (Cloudflare). Cloudflare added the identifying marks of that crawler to a ready-made blocking rule that "is available to all customers, including our free customers". Perplexity disputed the report.
On any Cloudflare plan, a setting such as Block AI training bots can block Perplexity's requests, including the ones made when your buyers ask a question.
On one client's site we read Cloudflare's security log, the Security Events page. The Block AI training bots setting was blocking real requests, including fetches made when a user asked a question. Trust the CDN's security log over any test run from outside.
On a second B2B client's site we read the web server's own log for 24 hours. PerplexityBot made 7 requests and none was blocked. The access check, in order:
Allow PerplexityBot in robots.txt, and allow both PerplexityBot and Perplexity-User by user agent in the firewall.
Add Perplexity's published IP ranges to the allow rule.
Turn off Cloudflare's Bot Fight Mode and the Block AI training bots setting.
Wait up to 24 hours, then read the CDN's security log for the two agents by name.
From this article.
Once Perplexity can reach your pages, your own site is still at most one source among several.
In Agenzy's 2026 source study of 2,283,923 AI citations (which sources AI engines cite), the brand's own site was cited on Perplexity in 66.9% of projects. That is 89 of the 133 projects with Perplexity data, at a median rank of 8, meaning typically the eighth most-cited source in its category. The next question is which other pages Perplexity cites for your category.
The sources Perplexity cites
ChatGPT, Google AI Overviews and Perplexity share only 2 of their top 10 cited sources when answering the same questions about the same brand, according to Agenzy's 2026 source study (n=109 business projects). The 109 are the projects with at least 10 cited sources on each of the three engines.
Google AI Overviews takes 10.8% of its citations from user-generated content against Perplexity's 3.8%, while Perplexity takes 13.6% from news and media against AI Overviews' 5.2%, Agenzy's 2026 study of 2,283,923 AI citations found (n=137 and 133 business projects). The 137 and 133 are the projects with AI Overviews data and with Perplexity data.
Agenzy 2026 source study.
The source study covers 130 European businesses, and 104 of its 138 projects ran from Lithuania, so the exact shares will differ in a US category.
Reddit shows the difference between the engines most sharply, and the Reddit figures changed between September and October 2026. Reddit is the single most cited domain in Agenzy's 2026 AI source study, taking 2.1% of 2,283,923 citations and appearing in 92.8% of 138 business projects, with ChatGPT citing it 5.0x more heavily than Perplexity. That data ends on 7 September 2026.
In our October 2026 tracking data from two English-language projects, Perplexity retrieved Reddit in 13% and 45% of chats, and ChatGPT in 0.3% and 4.7%. Perplexity cited Reddit in 3.8% and 14.5% of those chats. For an English-language market act on the October reading, and check the date of any Reddit figure before you use it.
Peec AI, October 2026.
What Perplexity cites depends on the category, and the three projects behind the search counts show it. For the B2B sales-outsourcing company, the 60 most-cited URLs received 556 citations. 181 of those, 33%, went to comparison and pricing pages, which is the kind of page a B2B software buyer also reads. For the consumer brand, listicles took 59% of the citations among the 30 most-cited URLs outside its own site.
Peec AI data.
Your off-site list for Perplexity, meaning the other people's pages where your company needs to be named, comes from Perplexity's own citations in your category. Those are the comparison pages, directories, trade press and review sites it cites today.
Freshness and dates
Perplexity records a date for every page and lists its sources newest first. So where the buyer's question depends on the year, the year belongs in the title. It still cites pages that are years old, so the date alone does not decide it.
Ahrefs, in a study of 17 million citations, found that "Perplexity and ChatGPT order their in-text references from newest to oldest" (Ahrefs). In-text references are the numbered sources inside the answer. Perplexity's in-text references in that study average 1,166 days since publication, so the typical cited page is about three years old.
In the B2B sales-outsourcing project, 21 of the 60 most-cited URLs carried "2026" in the title. In the Lithuanian video-production project, 17 of its 30 most-cited URLs did, most of them directory price pages. In the consumer sports-equipment project, a gift-list post from 2010 was among the most cited. A buyer expects prices, benchmarks and "best X" shortlists to be for the current year.
Do not change the date without changing the content, because a reader can see the date.
Writing the passage Perplexity quotes
Write each section as one passage that answers one buyer question in the buyer's own words. Make it stand alone when it is cut out of the page. That means the company name and the product name in place of "we" and "it", and the category and the market written out.
The one rule specific to Perplexity is exact words. Perplexity scores each passage of a page on its own, compared with the query. Its index matches words, so a passage is found when it carries the words the buyer typed.
An example of the shape, built from one of our own lines. Before: "We make brands visible in AI search." After: "Agenzy is a GEO agency working with B2B companies in the US, UK and EU on how often ChatGPT, Perplexity and Google AI Overviews cite and name them." The second version says who, for whom and on which engines when it is cut from the page.
From this article.
The only controlled experiment that measured writing changes on the live Perplexity engine is the GEO paper (GEO: Generative Engine Optimization, KDD 2024). Its "subjective impression" score gives each cited source a share of the answer, and the shares add up to the whole answer. On that score, adding statistics lifted a page from 24.7 to 33.9, and adding quotations lifted it to 32.1 (Table 5).
For a page that was already the leading source among the five the engine was given (rank 1), adding cited sources reduced visibility by 30.3%. For a page in fifth place, adding sources, quotations or statistics raised visibility by 97.9% to 115.1% (Table 2). For a company absent from Perplexity's answers, the fifth-place figure is the one that applies.
No test of structured data covers Perplexity. The evidence that exists is from Ahrefs (Ahrefs). It tracked 1,885 pages that added schema code (structured data written in JSON-LD) and compared them with 4,000 similar pages that did not.
Google AI Overviews citations fell 4.6%, a small but statistically significant decline, and ChatGPT and AI Mode showed no change beyond normal variation. So on that evidence schema code does not earn citations, and we keep it for Google's rich results.
Measuring your Perplexity visibility
Measure Perplexity visibility as two shares over a fixed set of buyer prompts. One is the share of answers that cite your site, the other the share that name your company.
Run the prompt set on a schedule in one of Perplexity's modes and keep to it, because Pro Search and Deep Research draw on different numbers of sources. The prompt set stays fixed for the whole project, because a changed set measures different questions. The steps:
Write 30 to 50 of your buyers' questions in their words, with the category and the market in them, and none that name your brand.
Run them daily in a tracking tool that covers Perplexity; our comparison of AI visibility tools lists which ones do. Record, per prompt, whether your site was retrieved and cited and whether your company was named.
Compare your citations per retrieval with Peec AI's benchmark for Perplexity: 1.5 to 2.0, where ChatGPT runs at 2.0 and above. Citations per retrieval means how many times a page is cited for each time Perplexity reads it.
Export the cited-sources table every month; it is next month's off-site list.
Peec AI report example.
Further reading: our guide to ranking on ChatGPT covers the engine most B2B buyers use first.
FAQ
Does Perplexity have a ranking algorithm?
Yes. Perplexity retrieves and ranks pages inside its own index, since September 2026 with Photon, its in-house retrieval and ranking engine. A reader sees one answer with a few numbered sources that change each time the question is asked. The measure is the share of your buyer prompts in which your site is cited and your company named.
Is Perplexity worth tracking in the UK and the EU?
Yes, as the engine to add after ChatGPT and Google's AI answers. Statcounter put Perplexity at 5.44% of AI chatbot market share in Germany and 3.08% in Poland in August 2026 (Statcounter). Its October 2026 reading moved sharply upward worldwide, so check the current month for your market before you decide.
Does Perplexity use Google or Bing search results?
No, by Perplexity's own account. It says it runs its own index "covering hundreds of billions of webpages", crawled by PerplexityBot (Perplexity Search API announcement). Where Perplexity's citations overlap with Google's top results, the two systems have read the same web.
What do Perplexity's Pro and Research modes change for sources?
They change how many sources an answer draws on and which model writes it, so they change which sources you see. Perplexity's help centre says Pro Search works "across multiple sources" and Deep Research "performs dozens of searches automatically, reads hundreds of sources" (Perplexity help centre).
How does Perplexity differ from Claude's route through Brave Search?
The work is different, so the Perplexity checks above do not cover Claude. Claude searches only on some prompts (Anthropic, web search tool) and, when it does, cites from Brave Search's results (MERJ), so getting into Brave's index is a separate job. How to get cited by Claude covers it.
Can a new page be cited within days?
Yes. Perplexity-User fetches a page live when a user's question calls for it (Perplexity, bots), so a page published today can be read and cited today. A share across your prompt set moves over months. The user agents for both Perplexity bots are in our list of AI crawlers.
What should I do when the CDN log shows PerplexityBot blocked?
Fix it in the firewall, following the access check in the crawler access section of this article. Read the same log 24 hours later to confirm that the PerplexityBot and Perplexity-User rows show 200 (allowed) responses.
Does Perplexity cite YouTube or LinkedIn?
Rarely. In our September 2026 test across six Peec AI projects, Google AI Overviews cited YouTube and LinkedIn most often and Perplexity second most often. A manufacturer with 636 videos received 14 YouTube citations across those engines in four months. One retailer's comparison-test video, titled as the question its buyers ask, received 49. YouTube's overall share across engines is in our study of which sources AI engines cite.
How does a brand get named without its own URL being cited?
When a comparison page, directory entry or review that Perplexity cites names your company, the answer can name you without linking your site. In one consumer project, two independent reviews of the brand's product sat among the 30 most-cited URLs on Perplexity. How often the median brand is named in each engine's answers is what our 2026 AI search visibility study measures.
What is the Perplexity Publishers' Program and who does it pay?
A revenue-sharing programme for publishers (Perplexity, Publishers' Program). When Perplexity earns advertising revenue from an interaction that references a partner's content, the publisher earns a share. It pays publishers whose content Perplexity already cites, and the announcement describes no ranking or citation preference for members.
How do I measure Perplexity visibility without a tool?
Run your fixed prompt set by hand in one Perplexity mode. Log three columns per prompt: whether your site is cited, whether your company is named, and which sources appear. Perplexity's Search API filters results by domain with search_domain_filter, up to 20 domains, which tests whether a page of yours is retrievable for a query (Perplexity, search filters). Our GEO metrics guide defines citations per retrieval and the other measures.
Final thoughts
Ranking in Perplexity in 2026 means having your site cited and your company named in Perplexity's answers, measured as a share of a fixed set of buyer prompts. Perplexity's two bots have to reach your pages.
Your company has to be named in the pages Perplexity already cites for your category. Each of your own pages has to carry the buyer's words, and the year where the question has one, in a passage that stands alone.
Two actions today: type ten of your buyers' questions into Perplexity and note which sources it cites and whether it names you. Then ask your web developer to read the PerplexityBot and Perplexity-User rows in the CDN's security log. Emilis Zabilius, who wrote this article, reads replies on LinkedIn; send him what you find.
