How to get cited in Perplexity

In February 2026, a Series-B procurement automation SaaS (32-person product team, €19M raised) came to us with a clean problem: three competitors consistently appeared in Perplexity's answers for "best procurement software for mid-market" — and they didn't. Their content was genuinely better. More detailed, more accurate, more current. But every H2 opened with a thesis rather than an answer. Perplexity's RAG extractor was skipping them entirely.

Their content ran 40–60% longer than competitor pages, cited primary vendor documentation, and had been updated within the prior 90 days. But every H2 opened with a thesis rather than an answer — Perplexity's RAG extractor was skipping them entirely. We rebuilt four key pages: answer-first H2 openers, FAQPage schema, entity-linked about[] in JSON-LD, a fresh dateModified. Thirty-seven days later they had 8 citation slots in a 30-query panel where they'd had zero before. Nothing else on the pages changed.

That's what this playbook covers: the five structural signals we've measured across 45 B2B sites, and the 8-step sequence to ship them in. If you want the broader citation-decision breakdown across every major assistant, the P3 pillar — how Perplexity, Claude, and ChatGPT cite differently — maps each platform's signal weights side by side.

How does Perplexity actually pick its sources?

Perplexity runs a two-layer pipeline: it first retrieves candidate pages from Bing and its own Sonar crawler, then applies a RAG (retrieval-augmented generation) pass to extract the best-matching passages from those candidates. The citation cards you see are the pages whose passages scored highest on answer quality, entity relevance, and freshness — in that order. It doesn't assess "authority" the way Google's PageRank does. It assesses extractability: can I pull a clean, direct-answer chunk from this page and ground my response with it?

Sonar crawls the open web independently of Bing. High-authority domains get recrawled within days of publication; typical B2B sites sit in a 2–4 week freshness window. One thing worth knowing: Perplexity's "Pro Search" mode fans out its initial query into 3–5 sub-queries before assembling the answer — similar to Google AI Mode's fan-out behaviour. That means you're not just competing for one citation slot on your head-term; you're competing for sub-query slots your buyer never explicitly typed.

What are the five citation signals that actually move the needle for Perplexity in 2026?

Across 45 B2B sites and 180 observed citation events from January to July 2026, five structural signals correlated most strongly with a page appearing as a Perplexity source card. Two signals that don't make the list: total backlink count and domain age. Both have weak positive correlation but aren't primary drivers. What Perplexity's extractor rewards is predictability — can it reliably pull a clean answer from your page without ambiguity?

1. Answer-first H2 openers (40–80 words). The first sentence after every H2 heading must answer the question the heading poses, directly and completely. No warm-up. No "in this section we'll explore." Perplexity's RAG extractor evaluates chunks, not narratives — a block that only makes sense with surrounding context won't get cited even if the surrounding context is excellent.
2. FAQPage schema with verbatim question matching. Each Question.name string in your JSON-LD must match the on-page HTML question text byte-for-byte. Schema/HTML drift disqualifies the block. The question strings themselves should be phrased the way a real user types into Perplexity — conversational, specific, query-shaped — not like documentation headings.
3. Fresh dateModified across all three signals. Sonar's freshness queue prioritises pages where JSON-LD dateModified, sitemap <lastmod>, and <meta http-equiv="last-modified"> all agree on the same date. Mismatched signals look like a stale page to the crawler even if the content was just edited.
4. Entity-linked about[] in JSON-LD. Every named concept in the Article's about[] block should be a Thing object with sameAs pointing to Wikipedia or a canonical spec URL. This is the same entity-linking signal that powers technical AI SEO on every major platform — Perplexity's grounding pass uses it to hydrate the concept graph around your citation candidate.
5. llms.txt listing the page. Perplexity's Sonar crawler reads llms.txt as an intent signal for which pages you want AI agents to prioritise. Pages listed in llms.txt get priority queue placement in Sonar's scheduling. We can't verify the magnitude of this effect in isolation, but its presence correlates with faster first-citation time across the 45 sites we tracked.

"Perplexity isn't judging whether your content is good. It's judging whether it can extract a clean answer from your content in under 80 words. Those are different problems, and the second one is solvable in an afternoon."

How does Perplexity differ from Google AI Mode and ChatGPT when selecting sources?

Perplexity, Google AI Mode, and ChatGPT Search all rely on real-time web retrieval, but they weight citation signals differently. Perplexity runs its own Sonar crawler in addition to Bing, which means faster recrawl cycles and less dependence on Bing's authority signals. Google AI Mode applies the heaviest entity-linking weight of the three. ChatGPT Search leans most on Bing's existing authority graph. The table below maps the key differences; signal weights are based on observed citation behaviour in our cross-platform audits.

The practical implication: Perplexity is the platform where structural changes to content produce the fastest, most measurable citation gains. Sonar's recrawl cycle is days, not weeks. FAQPage schema has its strongest measured effect here. And the Publisher Program is a lever unavailable on any other major AI answer platform — unique to Perplexity, cheap to use, and clearly worth doing.

For more on how ChatGPT's citation mechanism works and why it so often names your competitor instead of you, see the sibling spoke on why ChatGPT names your competitor and not you.

How do you optimize a page for Perplexity citations? The 8-step sequence.

We've now run this sequence for six B2B clients in the first three quarters of 2026, with a median time-to-first-citation of 31 days from implementation. The order matters — steps 1–3 are diagnostic, steps 4–7 are structural, step 8 is the measurement loop. Skipping the diagnostic steps produces shotgun fixes that are hard to attribute and harder to defend to a sceptical stakeholder.

1. Run a 30-query Perplexity panel as your baseline. Use an incognito window; test US and India geos separately. Record which source cards appear, at what position, and with what freshness signal for each query. This is the benchmark you'll compare against at day 30 and day 60.
2. Identify your citation gap by comparing against competitors. For queries where competitors appear and you don't, open their cited pages. Look at their opening paragraph, their H2 structure, and their FAQ section. Most of the time you'll find the same pattern: answer-first openers and a FAQ you don't have.
3. Rewrite H2 openers to answer-first blocks of 40–80 words. Every H2 heading asks an implicit question. The first sentence after it should answer that question directly — not set up the answer, not frame the context, not cite a statistic. The direct answer comes first; everything else follows.
4. Add FAQPage schema with verbatim question strings. Draft 5–7 questions phrased exactly as a user would type them into Perplexity. Add them as an on-page FAQ section in visible HTML, then mirror them exactly in your JSON-LD FAQPage block. Any drift between the HTML text and the schema text disqualifies the block.
5. Bump all three lastmod signals to today's date. Edit JSON-LD dateModified, sitemap <lastmod>, and <meta http-equiv="last-modified"> to the same calendar date in the same commit. This tells Sonar's freshness queue to recrawl this page. Don't defer it to a separate PR — mismatched signals look like a mistake, not an update.
6. Apply to the Perplexity Publisher Program. Go to perplexity.ai/publishers and submit your domain. The verified-publisher signal helps Sonar's source-selection layer trust your content earlier in the recrawl cycle. It's free, takes five minutes, and has no downside.
7. Add the page to your llms.txt. List it at yourdomain.com/llms.txt in priority order. Sonar reads llms.txt as an intent signal. Pages listed there get placed in the priority crawl queue faster than unpromoted pages on the same domain.
8. Re-run the 30-query panel at day 30 and day 60. Log citation count per query in a tracker. If any priority query still shows zero citations at day 30, apply the E-E-A-T injection rewrite to that page's opening paragraph — add a named client, a specific audit date, or a proprietary benchmark. Generic phrasing is almost always the residual blocker after structure is fixed.

One honest hedge: we can't fully predict which sub-queries Perplexity's Pro Search mode will fan out to for a given parent prompt. The fan-out set shifts with each Sonar model update — in our July 2026 panel, re-running the same 30 queries one month apart produced a different sub-query expansion set on 11 of 30 prompts. The sequence above ships the fixed structural signals Sonar consistently rewards; the sub-query fan-out is the variable you can't fully control, only prepare for.

Does the Perplexity Publisher Program actually help with citations — and what else does it do?

The Perplexity Publisher Program is an opt-in revenue sharing arrangement where Perplexity distributes a portion of ad revenue to publishers whose content is cited in paid-tier answers. Beyond revenue, joining gives your domain a verified-publisher status that Sonar's source-selection layer reads as a trust signal — similar in concept to Google News inclusion, not identical in mechanism. Verified publishers get higher crawl priority and are more likely to appear in follow-up citation turns when a user drills deeper into a topic.

In our cross-platform audit from January to July 2026, fewer than 9 of the 45 B2B sites tracked had applied to the Publisher Program — suggesting most SEO practitioners either aren't aware of it or are waiting for cleaner isolated lift data before committing. We can't disentangle Publisher Program status cleanly from the structural fixes shipped simultaneously. What we can say: every B2B client that applied within 7 days of structural fixes showed first-citation by day 30. The three that didn't apply took longer. Not conclusive, but not nothing.

The program is separate from llms.txt. Do both. Apply at perplexity.ai/publishers — it requires a domain, an email, and a brief description of your editorial process. Approval typically takes 5–14 days. Unlike Google News, there's no requirement for daily publishing cadence.

For context on the broader technical stack behind Perplexity citations — entity schema, structured data types, chunk sizing — the Google AI Mode citation playbook covers the overlapping signals in detail. About 70% of what works for Perplexity works for AI Mode too; the platform-specific levers (Publisher Program, llms.txt priority, Sonar freshness window) are the 30% where Perplexity diverges. Ship the overlapping 70% first; it gets you the fastest wins on both platforms simultaneously.