How to Rank in AI Search (ChatGPT, Perplexity, AI Overviews)

Overview

AI search — ChatGPT, Perplexity, Google's AI Overviews — doesn't return ten blue links; it answers the question and cites a few sources. Ranking in it means being one of those cited sources. The good news: most of what works is good SEO done well — clear structure, factual content, and authority — plus a few habits tuned for how language models read and reuse the web. Here's what actually moves the needle.

What gets your content cited by AI

AI search engines work in two steps: they retrieve a handful of candidate sources (usually from a search index), then synthesize an answer citing a few. To be picked, do two things. First, answer directly and early — lead each section with the answer, then explain (the inverted pyramid), because models extract clean, self-contained statements far more readily than points buried under throat-clearing. Second, make it specific and verifiable: concrete numbers, dates, and named examples beat vague marketing copy, and they survive the model's habit of cross-checking claims across sources.

Then structure for parsing — descriptive, question-shaped headings, short paragraphs, lists, comparison tables, and a real FAQ — so a model can lift a tidy, attributable snippet. Authority and consensus break ties: being mentioned and linked across the web, and agreeing with other credible sources, makes a model more confident citing you. This is exactly where classic SEO's brand mentions and backlinks pay off again in AI search.

• Answer-first: lead with the answer, then explain
• Concrete, verifiable facts (numbers, dates, named examples)
• Question-shaped headings, lists, tables, a real FAQ
• Authority + consensus: mentions and links across the web

Be readable by machines (where most sites lose)

If a crawler can't read you, an AI can't cite you — and this is the most common failure. Keep important content in server-rendered HTML, not locked behind heavy client-side JS that crawlers may not execute; keep the site fast and accessible; add structured data (FAQ, Article, SoftwareApplication); and crucially, don't block the AI crawlers whose traffic you want. Check robots.txt for GPTBot, OAI-SearchBot, PerplexityBot, and Google-Extended — blocking them guarantees zero citations from those assistants.

Because these engines retrieve candidates from a search index, ranking in classic search is also one of the strongest AI-search levers — the two reinforce each other, so don't treat them as separate projects. An llms.txt file gives agents a clean map of your site, but it's a helper, not a ranking factor (see llms.txt for SaaS). Mailbase's pages are built this way on purpose — server-rendered, FAQ/Article structured data, clear headings, llms.txt index. Going deeper: GEO, getting cited by ChatGPT, and AI search for SaaS.

• Server-rendered HTML for key content (not JS-only)
• Structured data (FAQ, Article) + fast, accessible pages
• Don't block GPTBot / PerplexityBot / Google-Extended in robots.txt
• Rank in classic search too — engines retrieve from the index

Common Mistakes

• Treating llms.txt as a ranking shortcut — it's a map, not a signal.
• Burying the answer under marketing copy a model can't cleanly extract.
• Locking key content behind heavy client-side JS that crawlers can't read.
• Chasing tricks instead of authority — answer engines are built to ignore them.

Related guides

More on how to rank in ai search and the surrounding llm seo workflow:

FAQ

Related reading