How to rank in Perplexity

To rank in the Perplexity AI answer engine, a page first has to be reachable by PerplexityBot, the crawler Perplexity says is designed "to surface and link websites in search results on Perplexity" (opens in a new tab). Allow it in robots.txt and at your firewall, and keep your key facts in server-rendered text. Studies show that Perplexity cites more sources per answer than ChatGPT or Google's AI features (opens in a new tab), and that its citations overlap Google's top results more than other assistants' citations do (opens in a new tab).

Perplexity documents crawler access; what decides citation after that comes from outside studies. This guide keeps the two apart. It is one of a set on how AI assistants find and cite sources, and its sources were checked on September 27, 2026.

Scope and definitions

Perplexity is an AI answer engine: when a user asks a question, it can visit web pages and include links to them in its response (opens in a new tab). Work to be cited by engines like it is called answer engine optimization (AEO) or generative engine optimization (GEO). The same practice across search engines and assistants is covered in AI search optimization.

• PerplexityBot is the crawler that surfaces websites for Perplexity's search results (opens in a new tab). Perplexity-User is the agent that fetches a page when a user asks a question (opens in a new tab).
• Retrieval is finding and fetching documents from an index to answer a question. A page that is not retrieved cannot be cited as a source.
• A citation is a link to a source shown with the answer.
• Sonar is the name of Perplexity's API models. Several studies below, such as a test of 112 startups through the Sonar API (opens in a new tab), used the API rather than the consumer app. That matters: in an audit of ChatGPT and Gemini, the API and the consumer interface of the same product shared only 12% to 15% of cited domains (opens in a new tab) (preprint). No study we found makes the same comparison for Perplexity.

How Perplexity finds and cites pages

Two agents, two purposes

Perplexity's crawler documentation describes two agents (opens in a new tab) (official documentation):

Perplexity publishes the IP addresses of both agents as JSON files, one for PerplexityBot and one for Perplexity-User (opens in a new tab), so a firewall can allow them by address as well as by user agent.

Two practical consequences follow. Blocking PerplexityBot works against being found, since Perplexity recommends allowing it "to ensure your site appears in search results" (opens in a new tab). Blocking Perplexity-User in robots.txt is unlikely to stop user-requested fetches, since Perplexity says that agent generally ignores those rules.

What is not documented: Bing and other indexes

Whether Perplexity also draws on a third-party search index, such as Bing, is not documented. Perplexity's crawler documentation describes only its own two agents and names no outside index (opens in a new tab), and our research found no official statement either way. Treat any claim that "Perplexity uses Bing" as unverified.

Other assistants document more. OpenAI says ChatGPT search sometimes partners with other search providers, and it names Microsoft among them (opens in a new tab). Anthropic lists Brave Search as a web search subprocessor (opens in a new tab) for Claude. For Perplexity, the only documented route into its results is PerplexityBot.

How many sources it cites, and which kinds

Perplexity cites more pages per answer than the other systems measured:

• In an academic audit of 4,706 queries, Perplexity's Sonar model retrieved a median of about 14 pages per query (opens in a new tab), against 9 for Google's AI Overviews and 4 for GPT-4o search (preprint; July to September 2025).
• In a descriptive dataset of 602 prompts, Perplexity averaged 16.35 citations per prompt, against 12.06 for Google and 6.88 for ChatGPT (opens in a new tab). The same study found that each page Perplexity cited had less influence on the answer (a mean score of 0.06, against 0.27 for ChatGPT) (opens in a new tab) (preprint; the influence score is the authors' own measure).
• Perplexity cited the least popular domains of the engines in the 4,706-query audit, with a median popularity rank of 5,647 (opens in a new tab).

The kinds of sources differ from other engines:

• In API tests with ranking-style prompts, 67% to 73% of the domains Perplexity cited were earned media (third-party reviews and editorial coverage), 17.5% to 23.8% were social platforms, and about 9% were the brands' own sites (opens in a new tab) (preprint; mid-2025). Perplexity's social share was the highest of the four engines tested; ChatGPT's was zero.
• A vendor reported that Reddit made up 46.7% of Perplexity's top 10 cited sources (opens in a new tab) from August 2024 to June 2025. That is a share of the top-10 list, not of all citations, and the vendor's prompt set is not public. Treat it as contested.

How citations relate to search rankings

Two vendor studies agree that Perplexity's citations overlap Google's results more than other assistants' citations do, though they disagree on how much:

• On 15,000 long-tail prompts, 28.6% of Perplexity's citations ranked in Google's top 10 for the original prompt, against 8.0% of ChatGPT's in-text citations and about 8% for the other assistants (opens in a new tab) (July 2025).
• On 5,000 keywords, Perplexity's citations showed more than 91% domain overlap and 82% URL overlap with Google's top 10 (opens in a new tab) (2025).

The first study used full prompts and the second short keywords, which likely explains the gap. Neither shows which index Perplexity uses, or that ranking causes citation. Both are consistent with ordinary search rankings mattering at least as much for Perplexity as for other assistants.

Freshness

In a vendor study of about 17 million cited URLs, AI assistants cited pages about 25% fresher than Google's organic results, and Perplexity's citations averaged 1,166 days old (opens in a new tab). That is younger than organic results (1,416 days in the study's table) but older than the averages for ChatGPT, Gemini, and Copilot (opens in a new tab). The same study observed that Perplexity orders its in-text references from newest to oldest (opens in a new tab). Age was measured from when the vendor's crawler first saw each URL, and the topics of the prompts were not controlled.

A laboratory study points to a mechanism. When researchers added artificial publication dates to passages, seven LLM rerankers promoted the passages given newer dates, moving the top 10's mean publication year forward by up to 4.78 years (opens in a new tab) (peer-reviewed). Perplexity was not tested.

Accuracy

Perplexity does not always describe its sources correctly. In a study of news answers by 22 public-service media organizations, 30% of Perplexity's answers had a significant issue and 15% had a significant sourcing issue (opens in a new tab), the lowest overall rate of the four assistants tested (May to June 2025). When asked to identify the source of news excerpts, Perplexity answered 37% of queries incorrectly (opens in a new tab) (1,600 queries across eight tools; March 2025). Both studies used news, not B2B topics.

Stability

Answers change between runs. In daily tests of ChatGPT, Gemini, Google AI Mode, and Perplexity, repeated runs shared only 32% to 43% of their cited sources, so about 65% of sources turned over from one day to the next (opens in a new tab) (preprint; German-language prompts; first author affiliated with a tracking vendor). One answer is not a measurement.

New and small brands

Perplexity surfaces new brands more often than a model answering from memory, but still rarely. Newly launched startups appeared in 8.29% of Perplexity's answers to discovery-style questions, against 3.32% for a model without web search (opens in a new tab) (preprint; master's thesis; 112 products; December 2025). In the same study, Perplexity visibility correlated with referring domains (r = 0.319) and with Reddit presence (r = 0.395), while a score of on-page GEO features showed no significant correlation (opens in a new tab). These are simple correlations on small samples.

Across engines including Perplexity, on a first answer to a generic prompt, household brands appeared 73% of the time, mid-market brands 44%, and small brands 11% (opens in a new tab) (preprint; the author co-founded the tracking platform studied).

What content changes have been tested

The often-quoted figure that rewriting a page raises its visibility in AI answers by 40% comes from one experiment, and a critical survey of 45 studies rejects it as a general claim (opens in a new tab) (preprint). In the original GEO experiments, adding quotations raised a source's share of the generated answer by about 41% and adding statistics by about 31%, in a simulated engine where the page was already retrieved (opens in a new tab) (peer-reviewed). The same paper's small Perplexity test, with sources supplied as uploaded files rather than found by search, found that keyword stuffing did worse than making no change on the paper's main measure (opens in a new tab).

Later work is less encouraging. A 2025 benchmark found statistically significant gains for conversational-SEO methods in 3 of 54 cases, and adding statistics lowered rank in 19 of 24 (opens in a new tab) (peer-reviewed). In a 2026 KDD study, rewriting page text to optimize it for generative engines hurt retrieval (opens in a new tab) (peer-reviewed). The same study found that placing the answer early in a document earned higher reranking scores (opens in a new tab).

Manipulation also works in the lab, which is not a reason to use it. Hidden instructions on product pages moved promoted products up almost three rank positions on Perplexity (opens in a new tab), and similar attacks worked on Perplexity and Bing Copilot in a separate study (opens in a new tab) (both peer-reviewed).

How much traffic Perplexity sends

In one B2B referral panel, Perplexity's share of AI referrals rose from 3.1% to 7.3% (opens in a new tab) between mid-2025 and March to April 2026 (vendor panel). Perplexity also crawls relatively little for each visit it sends: in July 2025, it made 195 page requests per referral, against 1,091 for OpenAI and 38,065 for Anthropic (opens in a new tab) (network measurement). Cloudflare notes that referrals from native apps may carry no referrer header, so the ratios may be overstated by an unknown amount (opens in a new tab).

Steps

Steps 1 to 3 follow Perplexity's documentation. Steps 4 to 9 are recommendations based on the evidence above.

1. Allow PerplexityBot in robots.txt. A crawler obeys the most specific group that names it and falls back to the * group only when no group matches (opens in a new tab). If you add a User-agent: PerplexityBot group, it will ignore your * rules, so repeat any Disallow lines you still need in it.
2. Allow Perplexity's IP ranges at your firewall or CDN. Perplexity recommends permitting requests from its published IP ranges (opens in a new tab). Then check your server logs: PerplexityBot requests should receive HTTP 200.
3. Expect Perplexity-User to fetch pages when users ask about them. Robots.txt will generally not stop it (opens in a new tab).
4. Put key facts in server-rendered HTML. In Vercel's network data from December 2024, none of the major AI crawlers rendered JavaScript (opens in a new tab). In a practitioner test, no AI system read facts that existed only in JSON-LD when it fetched a live page, and Perplexity answered from its stored index (opens in a new tab) (one test page).
5. Answer the question in the first lines, with specifics. State prices, dates, and specifications plainly. In 252,000 controlled trials across six models, explicit prices and recent dates raised the odds of being cited first in all six, while formatting alone had no consistent effect (opens in a new tab) (peer-reviewed; laboratory; Perplexity not tested; authors employed by a software vendor).
6. Rank in ordinary search for the buyer's questions. Perplexity's citations overlap search rankings more than other assistants' do, as shown above.
7. Update pages when facts change, with accurate visible dates. We found no evidence that changing a date without changing the content helps, and a false date misleads readers.
8. Earn third-party coverage. Seek independent reviews, editorial coverage, and genuine participation in the communities where your buyers ask questions. For how mentions relate to AI answers, see brand mentions and AI recommendations.
9. Check what Perplexity says about you, and measure over time. Ask it about your company and compare the answer with your site. Run a fixed set of buyer questions several times over several weeks. For analytics, note that GA4's default AI Assistant channel does not name Perplexity (opens in a new tab), so build a custom channel for perplexity.ai; Microsoft Clarity's AI channel does include Perplexity (opens in a new tab). Methods are covered in how to measure AI visibility and how to track AI referral traffic.

What the evidence shows and does not show

Antipatterns

An antipattern is a practice that looks helpful but fails or backfires.

1. Blocking PerplexityBot and expecting citations. Perplexity recommends allowing it for a site to appear in its results (opens in a new tab). Instead, allow it in robots.txt and at the firewall.
2. Relying on robots.txt to keep Perplexity-User out. It generally ignores robots.txt (opens in a new tab). Instead, decide what you publish on the assumption that it can be fetched.
3. Optimizing for Bing because "Perplexity uses Bing." That is not documented (opens in a new tab). Instead, make sure PerplexityBot can reach you, and do ordinary SEO for all search engines.
4. Hidden instructions or prompt injections. They moved rankings on Perplexity in the lab (opens in a new tab), but they deceive users, and Google's spam policies cover attempts to manipulate generative AI responses (opens in a new tab), which puts the same pages at risk in Google. Instead, state true facts plainly.
5. Keyword stuffing. It did worse than no change on the main measure in the one Perplexity test published (opens in a new tab). Instead, use the buyer's own words once, where they fit.
6. Rewriting every page in "GEO style." Rewrites rarely helped and often hurt in a controlled benchmark (opens in a new tab), and body rewrites hurt retrieval (opens in a new tab). Instead, improve relevance and add specific facts.
7. Judging visibility from one answer. About 65% of cited sources turned over from day to day (opens in a new tab). Instead, repeat the same questions over several weeks.

Checklist

FAQ

Does Perplexity use Bing or Google?

We could not verify that it uses either. Perplexity's crawler documentation describes only its own crawlers (opens in a new tab), and we found no official statement that it uses Bing or Google. Its citations do overlap Google's top results more than other assistants' citations (opens in a new tab), but overlap does not show which index it uses.

Does allowing PerplexityBot let Perplexity train AI models on my content?

Perplexity says PerplexityBot "is not used to crawl content for AI foundation models" (opens in a new tab).

Can I submit my site to Perplexity?

Perplexity's crawler documentation does not describe a submission form (opens in a new tab). It asks site owners to allow PerplexityBot in robots.txt and at the firewall. IndexNow, a protocol for notifying search engines of new URLs, lists Bing, Yandex, Naver, Seznam.cz, Yep, and Amazon as participants (opens in a new tab); Perplexity is not listed.

Is generative engine optimization for Perplexity different from SEO?

Mostly not. The documented requirement is crawler access (opens in a new tab). The evidence favors crawlable, specific pages that rank in ordinary search and have independent coverage. It does not support rewriting tricks (opens in a new tab).

How long does it take to appear in Perplexity?

No study has measured it. Claims that Perplexity picks up new pages within hours to days circulate in vendor blogs, but we found no primary data behind them.

Change log

• September 27, 2026: First published.
• September 27, 2026: Corrected the age of Google's organic results in the Ahrefs freshness study to 1,416 days, the figure in the study's table.

Next steps

To check whether AI crawlers can reach your pages and what AI assistants say about your market, AEO HQ's Instant AEO Audit ($499) is an automated audit of your site and up to three competitors. It includes a live crawl of your key pages, AI crawler access checks, and a look at how AI assistants answer your buyers' questions.