Perplexity is the AI search engine you most want to understand, because it's the only one that shows its hand: under every answer it lists the sources it cited, numbered and linked. That means two things. One, you can see exactly who's winning in your category. And two, showing up there isn't magic or FOMO — it's a game with rules you can read. Here they are.
Why Perplexity isn't “ChatGPT with sources”
The starting mistake is treating it like ChatGPT. Perplexity doesn't answer from memory: on every question it runs a live search, retrieves a handful of pages and writes the answer from them, citing each with a little number. ChatGPT, unless it's browsing, mostly draws on what it learned in training. The practical consequence is big: on Perplexity, what you shipped yesterday can be cited today; on ChatGPT, your content takes months to “land” — if it lands at all.
That changes the strategy. On Perplexity you're not fighting to get into the model's weights: you're fighting to be one of the five to ten pages its search retrieves and deems worth citing for that specific question. It's closer to plain old SEO than anyone likes to admit — but with an AI layer on top deciding who gets cited and who gets left out.
How Perplexity decides who to cite
Under the hood, Perplexity works like a RAG system: retrieve, rank, generate. It retrieves with hybrid search (BM25-style keywords + semantic embeddings), runs the candidates through several layers of relevance and quality reranking, and only then writes the answer citing the sources that survived the filter. You don't need the jargon; you need to know what that filter rewards.
Which pages it actually cites (and what that tells you)
Studies on Perplexity's real behavior converge on a pattern that's uncomfortable for brands. A Semrush analysis of over 230,000 prompts put Reddit, Wikipedia, LinkedIn, YouTube and Forbes among the most-cited domains, and found that around 60% of Perplexity's citations overlap with Google's top 10 organic results. AI search hasn't buried SEO: it leans on it.
- Classic SEO is still your base. If you're not in Google's top 10 for your topic, you start at a disadvantage: that top 10 is a big chunk of what Perplexity ends up looking at.
- Reddit and forums carry weight. Perplexity cites real conversations. Being present — honestly, contributing, not spamming — in your category's threads is a channel, not a waste of time.
- Comparisons and “best X” lists win. When someone asks “what's the best…?”, Perplexity looks for pages that already answer exactly that. If yours doesn't exist, they cite someone else's.
The levers that actually move the needle
If you want Perplexity to cite you there aren't ten tricks: there are five things that work and a lot of people repeating filler. Here are the five.
- Answer in the first 100 words. Analysis of the most-cited pages found 90% answered the core question within the first 100 words. If your article warms up for three paragraphs before getting to the point, Perplexity flags it as low-density and moves on. Front-load the answer.
- Keep the page fresh. Freshness is one of the strongest signals: an Ahrefs study of 17 million citations across AI search engines found that regularly updated pages get cited noticeably more than ones frozen for months. Show a visible date and revisit your key pieces.
- Mark up content with schema. Pages with valid structured data (FAQPage, HowTo, QAPage) show up 20% to 30% more in AI summaries, per industry counts. It's an afternoon's work and it buys you an edge.
- Build source credibility. Authority matters — trusted domains get cited more — but it's not just Domain Rating: it's clear authorship, verifiable data, and others citing you. Be the source, not the echo.
- Cover the cheap flank: llms.txt. Perplexity is one of the few engines that retrieves and uses the llms.txt file to prioritize which pages it looks at. It's no miracle, but here it earns the few minutes it takes to ship.
The mistake almost everyone makes
Optimizing for Perplexity “blind.” Since it cites its sources in plain sight, you have no excuse: ask the question that matters to you, look at which five pages get cited, and study them. That's your real competition, not the one you imagine. And measure it over time: if you ship or update something, watch whether you break into those citations over the following weeks. In GEO there's no official dashboard; you build your own.