How to Rank in Perplexity: Optimizing for an Index Nobody Else Crawls

Perplexity work sits inside a GEO engagement as a platform workstream, never as a standalone product. Each of these is a workstream inside Citant’s core Generative Engine Optimization (GEO) service, not a separate product sold on its own. Content can pass retrieval and re-ranking and still fail to get the brand named in the final answer, a failure Citant calls a ghost citation. For a B2B SaaS marketing team that needs to know whether Perplexity names the brand at all before deciding what to change, AI visibility services report Share of Model on each tracked platform separately. For teams that want Perplexity handled alongside the other five platforms rather than as a side project, our GEO service runs the 4-Layer Citation Framework across all six.

The two Perplexity user agents do two different jobs, and a robots.txt rule that treats them as one produces the wrong outcome twice. Perplexity documents PerplexityBot as the agent behind its search results, stating that it “is designed to surface and link websites in search results on Perplexity” and that it “is not used to crawl content for AI foundation models”. Perplexity documents Perplexity-User as a reader-triggered fetcher instead, stating that it “supports user actions within Perplexity” and that “since a user requested the fetch, this fetcher generally ignores robots.txt rules”. Source: Perplexity crawler documentation, fetched 23 August 2026. Last verified: August 2026.

Perplexity indexation is a separate technical job from Bing indexation and Google indexation, and no amount of work in the other two consoles substitutes for it. Perplexity publishes IP ranges for both of its agents, in a published PerplexityBot IP range file and a published Perplexity-User IP range file, so a publisher can confirm a real visit in server logs rather than trusting a user agent string. A firewall or CDN rule that blocks unrecognized bots stops PerplexityBot with no robots.txt evidence at all. Teams that want crawler access checked against a named list before rewriting anything can run the AI Crawler Access Report, a free tool from Citant.ai.

Token-level robots.txt syntax for every AI crawler belongs in one place on this site rather than being restated per platform. This page states only the PerplexityBot and Perplexity-User distinction, because that distinction is what changes a Perplexity decision. For the directive syntax, the full token list, and the vendor documentation dates behind each entry, the full AI crawler reference table carries the complete set across every named AI crawler.

A Perplexity citation is a numbered inline marker attached to a visible source list, not a link buried under a summary. Perplexity states that “each answer includes numbered citations linking to the original sources, allowing you to easily verify the information or explore further”, per the Perplexity Help Center, fetched 23 August 2026. Last verified: August 2026. That format sets the unit of work: the thing being cited is a passage a reader can check against the claim, so a page is scored by its passages rather than by its total length.

That range is a house drafting standard built on how retrieval-augmented systems split documents into retrieval units, and it is never presented as a measured citation multiplier. Every heading opens with a 40 to 75 word direct answer to the question the heading asks, before any setup. A passage lifted into a numbered source list carries no antecedent with it, which is why no factual sentence on a Perplexity-targeted page should open with an unanchored pronoun.

Perplexity overlaps Google’s top 10 more than the other tracked assistants do, and still nowhere near enough to treat one as a proxy for the other. In a University of Toronto study of 1,000 consumer ranking queries, the mean overlap between the domains Perplexity Sonar Pro cited and Google’s top 10 results was 15.2%, with a median of 14.3% (Chen et al., EDBT/ICDT 2026 Workshops), published as Navigating the Shift. Two bounds travel with that figure: the corpus is consumer ranking queries across ten retail and travel categories rather than B2B SaaS, and the paper reports no explicit calendar collection window. Read against a Perplexity plan, it says the same thing the crawler documentation says: a separate index needs separate work.

Perplexity spreads its citations across earned, brand-owned and social sources differently from every other assistant measured alongside it. In a University of Toronto preprint from the same research team, Perplexity allocated 73.4% of its citations for niche brands to earned sources, 9.1% to brand-owned domains and 17.5% to social sources (Chen et al., September 2025), published as Generative Engine Optimization.

Perplexity is one of only three engines in the measured literature whose citations have been audited against a page-level scoring framework on B2B SaaS pages specifically. A GEO-16 study collected 1,702 citations from 70 product-intent prompts and audited 1,100 unique URLs across Brave Summary, Google AI Overviews and Perplexity, finding that the pillars covering metadata and freshness, semantic HTML and structured data showed the strongest associations with citation (Kumar and Palkhouski, September 2025). Three cautions travel with that finding: the paper is a preprint with no peer review, it reports pillars that align with citation rather than cause it, and it publishes no uplift percentage. Its corpus is English-language B2B SaaS pages, which is the reason it is used here rather than a consumer study.

Paid placement and a cited source are two different objects on Perplexity, and conflating them is how a budget gets spent on the wrong one. Perplexity published its advertising position on 12 November 2024, stating that ads “will be formatted as sponsored follow-up questions and paid media positioned to the side of an answer”, that “the content of the answers you receive on Perplexity will not be influenced by advertisers”, and that “answers to Sponsored Questions will still be generated by our technology, and not written or edited by the brands sponsoring the questions”. Source: Perplexity, Why we’re experimenting with advertising, 12 November 2024, fetched 23 August 2026. Last verified: August 2026.

Nine steps, in the order the dependencies actually run. Nothing below step 3 changes an outcome until steps 1 through 3 are true.

Measurement on Perplexity is a naming test, run against a fixed query set, not a traffic report. Run each target query in a signed-out Perplexity session and record whether the brand is named in the answer text, which competitors are named instead, and which domains appear in the numbered source list. Repeat each query three times in the same session type, because one run measures variance rather than position. Log every result by query and by date, so the next pass measures movement instead of a single impression. Teams that want that baseline established for them rather than run in-house can start with an AI Search Audit to check Perplexity visibility.

No submission route to Perplexity’s index exists in its published crawler documentation, which covers allowing and blocking its two agents and publishes their IP ranges, and lists no inclusion request or indexing request for site owners. No purchase route to a citation exists either: sponsored placement, where a platform runs it, is a labeled advertising unit sitting beside an answer, not a source inside it. No agency controls whether Perplexity retrieves a given page, how many sources an answer shows, or which of them it names. Unlike guidance that sells a schema type or a robots.txt line as the Perplexity lever, the Perplexity workstream inside a GEO engagement works on the two things a publisher actually holds: crawler access and passage clarity.