Perplexity AI and SEO: how the engine picks sources and how to get cited

Perplexity does not return a list of links: it answers, then cites the pages it used. For a B2B brand, SEO on Perplexity therefore means getting cited in those answers, and knowing how to check it. This guide covers how the engine picks its sources and how to get cited, limited to what is specific to Perplexity: source selection, crawlers, modes, what it appears to favour and measurement. The general principles are covered in our guide to Generative Engine Optimization (GEO).

Perplexity at a Glance

Perplexity presents itself as an answer engine: for each question, it runs a real-time web search, reads the pages it finds, writes a synthesis and displays the sources used. The conversation keeps context from one question to the next.

Usage remains a minority compared with Google, but it is far from marginal. The latest published volume dates from May 2025. At that point, Perplexity was handling 780 million queries per month (Affmaven, 2026).

For a brand, this changes the unit of visibility. There is no second page of results: your page is either cited in the answer or it is not. A citation counts even without a click, because the reader sees the source's name next to the claim.

How Perplexity Selects and Cites Its Sources

Perplexity publishes no official ranking factor. What follows describes the engine's visible behaviour at the time of writing, supplemented by what can be observed by re-running questions. Any observed pattern is correlation, not causation.

The Steps Behind an Answer

An answer goes through five steps:

• Query: the user's question, possibly reformulated into several searches.
• Search: real-time querying of the web and of Perplexity's own index.
• Selection: filtering of the pages judged relevant, usually 5 to 10 sources per answer.
• Synthesis: writing an answer that combines the passages retained.
• Citations: inline numbered citations attached to each claim, plus a list of sources below the answer.

Citation is part of the product itself: Perplexity includes citations in 92% of its answers (DataGlobeHub, 2026). Each number points to a specific page, so you know which passage was reused.

Why Sources Vary

Two users asking the same question do not necessarily get the same sources. At least three factors shift the list:

• Wording: a different word triggers different searches, and therefore different pages.
• Timing: the index changes continuously, and a page published yesterday can replace an older one.
• Mode: Pro Search or a restricted focus changes the number and type of sources consulted.

The consequence for measurement: a single check proves nothing. Only a question panel re-run at regular intervals shows whether a page is cited consistently or by chance.

Perplexity's Crawlers

Perplexity uses two distinct agents, which are easily confused in robots.txt. Their roles are described in Perplexity's documentation, at the time of writing.

PerplexityBot and Perplexity-User

The table compares the two agents on what matters to a brand:

Regarding the second agent, several investigations have reported access that may not be declared under these names. These observations, to be read with caution, are a reminder that robots.txt expresses a rule without any technical guarantee. To put these agents in context, see our article on AI agents and how they work.

Configuring robots.txt to Get Cited

If your goal is to be cited, blocking PerplexityBot removes your pages from answers. A minimal configuration allows the agent across the whole site and closes only private areas:

• User-agent: PerplexityBot
• Allow: /
• User-agent: *
• Disallow: /admin/
• Disallow: /customer-area/

Before making any change, review your existing rules: a blanket ban (User-agent: * followed by Disallow: /) also blocks PerplexityBot. The directives are covered in detail in our guide to the robots.txt file.

Perplexity's Modes and Surfaces

Perplexity is not limited to a single interface. Each mode consults different sources, which changes the list of pages that stand a chance of being cited.

Pro Search and Focus Modes

Pro Search runs a multi-step search: it breaks the question down, chains searches and consults more sources before answering. Complex B2B questions (comparing offers, assessing a method) are well suited to it.

Focus modes, for their part, restrict the search to one type of source: web (the default setting), academic, social, finance or video. A methodical study belongs in academic focus; in social focus, the brand appears through what others say about it.

Pages, Spaces, Comet, Shopping and Finance

The other surfaces widen the touchpoints between Perplexity and a B2B brand:

• Pages: pages generated and then published by users, which cite their sources like an answer and may reuse your content.
• Spaces: workspaces where a team gathers its research and documents.
• Comet: Perplexity's agentic browser, which visits websites and completes tasks on them for the user.
• Shopping: product cards, based on specific sources and mainly useful for offers sold online.
• Finance: financial data, relevant to a listed company or a financial services player.

For most B2B brands, web search remains the main battleground; the other surfaces are worth tracking if your business fits them.

Perplexity, Google (AI Overviews / AI Mode) and ChatGPT: Who Cites What

All three engines cite sources, but they do not look for them in the same place. At the time of writing, the difference lies mainly in the index used:

The direct consequence: being cited by one does not mean being cited by another. A page that ranks well on Google can be missing from Perplexity if PerplexityBot has not crawled it, and the reverse also happens. Each engine should therefore be tracked separately, using the same question panel.

The common foundation remains an accessible, clear and up-to-date site: see GEO vs SEO: what changes for visibility.

What Perplexity Favours, and How to Provide It

With no official factors, we can only talk about observed signals: the pages Perplexity cites regularly share a few traits. These are likely levers, not rules.

Observed Signals

The table links each signal to what you can produce:

Structure remains the best-documented signal: pages with an H1-H2-H3 hierarchy are 2.8× more likely to be cited (State of AI Search, 2025). This figure, across all engines, describes a correlation.

The Format of a Page That Answers

A page that is a candidate for citation combines:

• a direct answer at the top of each section, reusable as is;
• short sections, one idea per paragraph, headings that restate the question;
• lists for steps and criteria, tables for comparisons;
• sourced and dated data, with its scope;
• a visible update date and author.

These rules belong to an overall editorial strategy, described in our guide to GEO content strategy. For the writing itself, the principles of SEO copywriting remain the basis.

What Has Not Been Proven

Several recommendations circulate without evidence. At the time of writing, none of them rests on information published by Perplexity:

• Structured data: it describes the page to engines, but no direct effect on Perplexity citations has been established (see structured data and GEO).
• The llms.txt file: it offers models a map of the site, with no demonstrated effect on citations (see our llms.txt guide).
• Perplexity "cannibalising" Google: no data shows that optimising for one costs you on the other; both rely on accessible, clear and up-to-date pages.
• "Perplexity factors": the weighted criteria lists in circulation are reconstructions, not published rules.

The right stance is to test: change one group of pages, keep a control group, re-run the question panel and compare.

Measuring Your Presence on Perplexity

Presence on Perplexity is measured at two levels: the traffic that citations send, and presence in answers, with or without a click.

The perplexity.ai Referrer in GA4

When a reader clicks a citation, the visit arrives with the perplexity.ai referrer. In Google Analytics 4, it appears in the acquisition reports under the source / medium dimension; a dedicated segment lets you track sessions, landing pages and conversions. Some visits may arrive without a referrer, so the volume you see is a floor.

The Question Panel and Share of Voice

For presence in answers, build a question panel:

• 20 to 50 real questions from your buyers, from discovery to decision;
• re-run at regular intervals, under the same mode and language conditions;
• scored for each answer: brand presence, citation position, accuracy of what is said;
• supplemented by the sources cited instead of yours, to calculate share of voice against competitors.

Whether recorded by hand or with an AI visibility tracking tool, the panel feeds these indicators:

Read these indicators by page group (guides, offer pages, documentation) rather than URL by URL: source variability makes individual movements unreliable. The full approach is described in our guide to the GEO audit for AI engines; to put your results in the context of the engine's overall usage, see our Perplexity statistics.

A 4-Step Method to Get Cited by Perplexity

The 4 Steps

The method relies on a short cycle, to be repeated:

• 1. Question panel: select 20 to 50 high-stakes questions and record who is cited today.
• 2. Audit of cited pages: analyse your cited pages and those of competitors (direct answer, structure, data, date, author).
• 3. Production or rewriting: first rework existing pages into the answer format, then create the ones that are missing.
• 4. Measurement and iteration: re-run the panel, compare with the initial baseline, adjust.

In Incremys, the GEO & SEO platform, the 360° GEO & SEO audit tracks AI visibility by keyword, Perplexity included; the content production and performance reporting modules support steps 3 and 4.

Depending on Your Situation

Your starting point depends on what you already have in place:

• Your visibility still relies on Google alone: start by moving from SEO to GEO.
• You need to convince internally: draw on our GEO statistics.
• Your site is marked up or soon will be: follow our guide to structured data, without expecting a direct effect on Perplexity.
• You are already cited but poorly described: prioritise accuracy in the panel and correct the source pages.

FAQ on Perplexity and SEO

Should you block PerplexityBot?

No, if you want to be cited. PerplexityBot feeds Perplexity's search index: blocking it in robots.txt removes your pages from answers. Blocking Perplexity-User has a less certain effect, since Perplexity states that this agent may ignore robots.txt for requests triggered by a user.

How many sources does Perplexity cite?

Usually 5 to 10 sources per answer, shown as numbered citations in the text and as a list below the answer. The number varies with the question and the mode: Pro Search consults more sources than a quick search. Perplexity publishes no fixed rule; this range describes what can be observed at the time of writing.

Will Perplexity replace Google?

No. Its market share remains small compared with Google, and the use cases differ: Google is for browsing results, Perplexity for getting a sourced answer, particularly for research and comparison. The two complement each other: a clear, up-to-date page accessible to both crawlers can be cited by either, with no cross-guarantee.

How do you measure traffic from Perplexity?

Through the perplexity.ai referrer in Google Analytics 4: filter the source / medium dimension to track sessions, landing pages and conversions. This volume only covers clicks on citations, and some visits arrive without a referrer. Complement it with a question panel re-run regularly, which measures presence in answers, with or without a click.

Why do the sources cited change?

Because Perplexity runs a new search for every question. The wording, the timing and the chosen mode change the pages found: a more recent or more direct competing page can take your place, then lose it. Hence the value of a question panel tracked over time and read by page group.

Does structured data help on Perplexity?

No direct effect has been proven at the time of writing. Structured data remains useful for traditional search engines and does no harm. For Perplexity, what emerges from observation lies elsewhere: explicit headings, a direct answer, lists and tables, sourced facts and a visible update date.