AI SEO: how to get cited by ChatGPT, Claude, Perplexity & Google AI

Last updated 2026-06-10

AI assistants increasingly answer questions directly and cite a handful of sources instead of returning ten blue links. AI SEO — also called GEO (Generative Engine Optimization) or AEO (Answer Engine Optimization) — is how you become one of those cited sources. The mechanics are concrete and mostly under your control. This guide walks the four levers, in order of impact.

How AI answer engines pick which sources to cite

Whether it is ChatGPT search, Perplexity, Claude, or a Google AI Overview, the selection pipeline is roughly the same: a crawler indexes the open web, a retrieval step pulls candidate passages for a query, and the model synthesizes an answer and attributes the passages it leaned on. To be cited you must clear every stage — be crawlable, be retrievable (extractable + relevant), and be trustworthy enough to attribute. Miss any one and you are invisible no matter how good the content is.

Lever 1 — Let the right AI crawlers in

This is the prerequisite that everything else depends on. If the crawler can't fetch the page, none of the rest matters. Each major assistant uses distinct user-agents for distinct jobs — and the most common, costly mistake is blocking a search crawler when you meant to opt out of training. For example, you can allow OAI-SearchBot (ChatGPT search) and PerplexityBot (to be cited) while separately deciding on GPTBot (OpenAI training). Google splits this cleanly: Googlebot handles Search and AI Overviews, while the Google-Extended token controls Gemini training without touching your ranking.

See the full breakdown — user-agents, purpose, and copy-paste robots.txt rules — in the AI crawlers reference. If you want citations and AI-search traffic, allow at minimum the search/answer crawlers: OAI-SearchBot, PerplexityBot, Googlebot, and the on-demand fetchers ChatGPT-User + Claude-User.

Lever 2 — Be in the indexes AI engines actually retrieve from

Crawl access is necessary but not sufficient — each answer engine retrieves candidates from a specific index, and you have to be in it. The one most sites miss: ChatGPT search retrieves web results via Microsoft Bing's index(OpenAI's search partner), alongside its own OAI-SearchBot crawling. A page that ranks #1 on Google but is thin or absent in Bing is largely invisible to ChatGPT.

• ChatGPT → Bing's index. Check site:yourdomain.com on bing.com, register at Bing Webmaster Tools (one-click import of your verified Google Search Console property), submit your sitemap.
• Google AI Overviews / Gemini → Google's index. Standard Search Console + sitemap discipline covers you.
• Perplexity → its own index built by PerplexityBot — allowing the crawler (Lever 1) is the coverage step.
• Speed up both Bing and others with IndexNow pings on every publish — new URLs get indexed in hours instead of weeks.

Lever 3 — Make your content extractable

Retrieval favors content a machine can lift cleanly. Concretely: serve real HTML (most AI crawlers do not execute JavaScript, so client-only-rendered content is invisible to them); write quote-ready headings and a tight one-sentence answer near the top of each page; add schema.org structured data (Article, FAQPage, BreadcrumbList, and a Speakable block for voice); and keep one clear topic per URL. A short FAQ that mirrors real questions is one of the most reliably-cited formats because each answer is a self-contained, attributable passage.

Lever 4 — Be corroborated, fresh, and credible

Models prefer to cite claims they can corroborate. Facts that appear consistently across your site and independent sources are safer to attribute, so being referenced elsewhere compounds. Freshness matters too: keep a visible and structured dateModified, and update cornerstone pages on a cadence. And credibility — a clear author/organization identity, an about/methodology page, and outbound citations to primary sources — raises the odds a model trusts you enough to name you.

Where does llms.txt fit? (honest answer: optional)

llms.txt is a markdown file at your site root that lists the pages you most want cited, your crawler policy, and your attribution preferences. The honest 2026 status: no major AI provider has confirmed reading it, and Google has said it doesn't use it. It is not a citation lever today — the four levers above are. It IS a zero-cost forward bet on an emerging convention: if adoption comes, early publishers benefit; if not, you lost five minutes. If you publish one, keep it consistent with robots.txt — robots.txt is access control, llms.txt is stated intent, and they should never contradict each other. The free generator builds a spec-compliant file whose permitted/restricted sections mirror your robots.txt rules; the validatorscores any live site's file against the spec.

The practical AI SEO checklist

• Allow the search/answer crawlers you want citations from (see the crawler reference).
• Register in Bing Webmaster Tools + submit your sitemap — ChatGPT retrieves via Bing's index.
• Decide your training-crawler policy deliberately (GPTBot, ClaudeBot, CCBot) — separate from search.
• Serve real HTML; never gate primary content behind JavaScript.
• Lead each page with a tight, quote-ready answer; use clear H2s.
• Add Article + FAQPage + Speakable schema.
• Set and surface dateModified; refresh cornerstone pages.
• Establish author/organization identity + methodology for credibility.
• Optional: publish an llms.txt (cheap forward bet — see the honest status above), validate its score, and benchmark against the directory.

FAQ