If you’ve been optimising for Google for years and assumed Perplexity works the same way, you’re leaving citations on the table. Perplexity SEO optimisation — the practice of structuring content so it gets cited in Perplexity AI answers — follows its own logic, and it differs from Google search in ways that matter. At Choco Media, we’ve been studying which content patterns surface in Perplexity answers and which get skipped over, and this post documents what we’ve found.
This is a practical guide for content teams, SEOs, and anyone who wants their site to appear inside AI-generated answers — not just on a results page. If you’re already familiar with GEO (Generative Engine Optimisation) at a high level, we’ll get into the specifics that most introductory posts don’t cover. If this is your first encounter with the topic, we have a primer on GEO vs. classic SEO worth reading alongside this one.
By the end you’ll know: how Perplexity selects sources, the structural patterns that improve your chances, and the practical steps to apply this week — not in three months after a site redesign.
How Perplexity actually selects its sources
Perplexity isn’t a search engine that returns pages. It’s a retrieval-augmented generation system — it retrieves documents, extracts relevant passages, synthesises them into a response, and cites the sources it used. The selection happens in two phases.
First, Perplexity runs a web search against its index (Bing-powered, with its own crawler on top). It ranks candidate pages using signals similar to traditional search: domain authority, freshness, relevance. Second, it reads those pages and extracts passages that directly answer the query. The pages that get cited are the ones that both rank well enough to be retrieved and contain prose that maps cleanly onto the question being asked.
- A page can rank highly but still get skipped if its answer is buried in a wall of generic text
- A page with modest traditional SEO strength can get cited if it answers a specific question clearly and early
- Pages with good structured data are more parseable — Perplexity respects FAQ, HowTo, and Article schema
- Freshness matters more than it does in Google for time-sensitive queries; Perplexity tends to prefer sources from the last 12 months
Understanding this two-phase selection is the foundation. Most optimisation advice focuses only on the first phase (getting into the index). The second phase — getting extracted — is where the content structure decisions pay off.
Answer-first writing: the single most reliable pattern
Perplexity’s extraction layer looks for dense, direct answers near the top of a document. The pattern that works consistently is sometimes called “answer-first” or BLUF (Bottom Line Up Front) writing: state the answer, then support it.
In practice this means the first sentence of any section answers the question implied by the heading. Not “in this section we’ll explore…” — that’s throat-clearing. The extraction model doesn’t care about your narrative arc; it wants a string of words that maps to the query.
What answer-first looks like in practice
Compare these two opening sentences for a section on email open rates:
- Buried answer: “Email marketing has been around for decades and remains one of the most effective channels. Many factors influence how recipients engage with messages…”
- Answer-first: “Average email open rates across industries sit between 20–25% in 2026, though B2B newsletters consistently outperform this at 30–35% (Mailchimp benchmarks).”
The second version is what Perplexity can extract and cite. It’s specific, it answers the implied question (what are typical email open rates?), and it contains a source attribution — a detail Perplexity’s system appears to weight positively when forming citations.
This applies at every level: article opening, section opening, even list item construction. Lead with the finding, then explain it.
FAQ blocks and their outsized role in Perplexity citations
FAQ sections are cited disproportionately often in Perplexity answers. This is partly because they’re structured as question-and-answer pairs — which maps directly onto the format of a user query. It’s partly because most FAQ sections use plain HTML that’s easy to parse. And it’s partly because good FAQ content tends to be dense: every sentence carries information.
How to write FAQ blocks that get cited
- Write questions in the same natural language a user would type — not “What is the function of X?” but “How does X work in practice?”
- Keep answers to 2–4 sentences. Longer answers get truncated in extraction; shorter ones lack enough signal.
- Include one specific fact, figure, or named process in each answer — vague answers don’t extract well
- Mark up the section with FAQPage schema (JSON-LD). Perplexity reads schema and it improves parse accuracy.
- Place the FAQ section mid-page or near the end — not as a footer afterthought. It should read as genuine content, not SEO filler.
We’ve written a detailed technical walkthrough on FAQ schema for AI Overviews and voice search that covers the JSON-LD implementation. The structural logic applies directly to Perplexity as well.
“Perplexity users ask real questions and expect answers that read like they were written by someone who knows the subject. If your page reads like a keyword-stuffed SEO document from 2018, it won’t get cited — regardless of your domain authority.”
The role of specificity and named sources
One of the clearest patterns in content that gets cited by Perplexity: specificity wins. Pages that make general claims (“AI tools can save marketing teams significant time”) are passed over in favour of pages that make specific ones (“We reduced brief preparation time from 45 minutes to 10 using a GPT-4o prompt chain — here’s the chain”).
This shows up at the level of:
- Statistics with sources: “68% of B2B buyers now conduct independent research before contacting a vendor (Gartner, 2025)” outperforms “most buyers research independently”
- Named tools and methods: Naming the specific tools, frameworks, or processes you’re describing — not generic category references
- Dates and version specificity: “as of June 2026” or “in Claude Sonnet 4” signals freshness and precision
- First-person observations: “In client work we’ve found…” or “we typically see…” — these signal original observation, which Perplexity weights as primary source material
The reason specificity matters at the extraction phase is straightforward: Perplexity is trying to give its user a useful answer. An extracted passage with a specific figure or named process is more useful than a vague claim. The algorithm reflects this.
How Perplexity differs from Google AI Overviews
If you’ve already done work for Google’s AI Overviews, some of it transfers — but not all of it.
Key differences to keep in mind
- Domain authority weight: Google AI Overviews lean heavily on established domains. Perplexity is somewhat more willing to cite smaller sites if the passage quality is high. This means there’s more room for newer or lower-authority sites to appear in Perplexity answers.
- Query scope: Perplexity handles more complex, multi-part queries than Google typically surfaces in AI Overviews. Content that covers a topic in depth — not just a single angle — tends to perform better.
- Commercial content tolerance: Google AI Overviews actively de-emphasise pages with heavy commercial intent. Perplexity is less aggressive about this, though pages that are mostly calls-to-action rather than genuine information still don’t get cited.
- Citation visibility: Perplexity shows citations prominently and users click them. Google AI Overviews citation UX is more buried. This means Perplexity citations can drive meaningful referral traffic — something that’s harder to measure from AIO appearances.
For a broader comparison of how AI search systems differ from traditional SEO, our post on AI SEO in 2026 covers the full landscape including ChatGPT, Gemini, and Google AI Overviews.
Technical signals that improve Perplexity citation likelihood
Structural and technical factors play a supporting role. They’re not as decisive as content quality, but they remove friction from the extraction process.
- Clean HTML: Perplexity’s crawler processes HTML. Pages with heavy JavaScript rendering, content hidden behind login walls, or complex DOM structures are harder to parse — and sometimes not indexed at all.
- Crawlability: Ensure your robots.txt doesn’t inadvertently block Perplexity’s bot (PerplexityBot). You can verify this in your server logs.
- Page speed: Slower pages are sometimes not fully crawled. Core Web Vitals improvements that you’d make for Google apply here too.
- Schema markup: Article, FAQPage, HowTo, and Speakable schema all help. Perplexity parses JSON-LD and it improves the accuracy of passage extraction.
- Canonical tags: If you have duplicate or near-duplicate content across URLs, canonical tags help Perplexity identify the authoritative version to cite.
None of these technical points are Perplexity-specific — they’re good technical SEO practice. The difference is that Perplexity’s extraction layer amplifies the cost of getting them wrong, because a page that can’t be cleanly parsed simply doesn’t get cited.
Content types that perform well in Perplexity
Not all content formats are equally likely to be cited. Based on patterns we’ve observed, these perform consistently well:
- Comparison and decision content: “X vs Y” and “how to choose between A and B” formats map cleanly onto the research queries Perplexity users typically ask. This content gets cited frequently because users come to Perplexity to make decisions.
- Process and how-to content: Step-by-step content with numbered lists is easy to extract. Each step is a discrete, parseable fact.
- Definition and explainer content: “What is X” content is heavily cited because it answers the definitional queries that make up a large share of Perplexity’s query volume.
- Data-backed opinion pieces: Posts that take a clear position and support it with specific data — not just “here are both sides” — tend to extract well because they contain dense, attributable claims.
Content types that underperform: purely narrative or storytelling content without factual anchors, sales-heavy landing pages, and content that’s structured for human engagement (dramatic openings, cliffhangers) rather than information density.
A practical audit: is your content Perplexity-ready?
Here’s a quick checklist to run on your highest-traffic posts before doing any new content creation:
- Does the article title match a natural question a user would type into Perplexity?
- Is the target question answered in the first 100 words?
- Does each H2 section open with a direct, specific statement rather than a transition or scene-setting sentence?
- Are there at least one or two specific statistics with named sources?
- Is there a FAQ section with 4–6 Q&A pairs using FAQPage schema?
- Are you naming tools, methods, or frameworks rather than describing them generically?
- Is the page cleanly crawlable — no login walls, no robots.txt block for PerplexityBot?
- Is the page date-stamped and updated in the last 12 months for time-sensitive topics?
If you score fewer than 5 out of 8 on your most important posts, content restructuring — not new posts — is the highest-leverage move right now. Our team is available for a brief audit call if you want a second pair of eyes on a specific page before you rewrite it.