Product
AI visibilityCompetitorsCitationsGEO score and page fixesPrompts and demandAI crawlers and referralsContent StudioAgents and automationsAurora strategistCommerceEarned mediaLocal networkAPI and MCPThe whole platform
Solutions
Brand and marketingSEO and GEO teamsContent teamsPR and communicationsE-commerceMulti-location and localAgenciesExecutives
Free tools
Free AI visibility scanPrompt set generatorAI readiness checkTemplates
Resources
Learn GEOUse casesHoliday AI Index 2026BlogNewsletterMethodologyDevelopersChangelogIntegrationsSecurity and trust
PricingAboutSign in
Check your AI visibilityBook a demo
How-to

How to Get Cited by ChatGPT & Perplexity: Three Gates

AI engines cite a page only if they can fetch it, choose it, and extract from it. A practical walkthrough of all three gates with the specific fixes for each failure mode.

By John Patterson, Founder · Published

The short answer

A page earns an AI citation by clearing three gates in order: the engine must be able to fetch it, must choose it from the candidate pool for a given question, and must be able to extract a clean answer from it. Each gate has distinct failure modes and fixes. Fetchable fails on robots.txt blocks, JavaScript-only rendering, and paywalls; fix with crawler access and server-rendered HTML. Chosen fails on vague titles, generic slugs, and format mismatch; fix by making the page's "cover" (title, URL, description, date) signal exactly which question it answers. Extractable fails on buried answers, accordion-hidden text, and image-only information; fix by putting a liftable 200-400 word answer at the top under a question-shaped heading. Diagnose in order: there's no point polishing extraction on a page the crawler can't reach.

Gate 1: Can the engine fetch your page?

AI crawlers are less capable than Googlebot and easier to block by accident. Work through this checklist:

  • robots.txt allows AI crawlers. GPTBot and OAI-SearchBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot, Google-Extended, and CCBot. Check for CDN-level bot blocking too: Cloudflare and similar services ship AI-bot blocking that overrides your own robots.txt, and it's often on by default. (We found exactly this on our own site: our robots.txt now explicitly welcomes every major AI crawler.)
  • Content is in the initial HTML. AI crawlers largely don't execute JavaScript. If your content renders client-side, the crawler sees an empty shell. Server-render or pre-render anything you want cited.
  • No auth walls or paywalls on pages you want cited, and no aggressive rate limiting that times crawlers out.
  • Help discovery: a sitemap.xml listed in robots.txt, and an llms.txt file that gives LLM agents a map of your site.

Symptom of a Gate 1 failure: you're rarely or never mentioned even for prompts where you should be an obvious answer, and your server logs show no AI-crawler visits.

Gate 2: Does the engine choose your page?

Before reading a page, retrieval systems judge its "cover": URL slug, title, description, and date. The cover must signal that this page answers this question.

  • Title precision. "Best AEO Tools in 2026: 8 Platforms Compared" beats "Our Thoughts on AI Search Tools." Include the category, the use case, and a freshness signal.
  • Slug matches topic. /learn/best-aeo-tools-2026, not /blog/post-847.
  • Format matches question shape. Comparative questions get answered from comparison-shaped sources. If the prompt is "X vs Y" or "best X", a single-product page loses to a structured comparison regardless of quality.
  • Visible dates. Publish and update dates on the page and in schema. Engines prefer sources that look maintained.

Symptom of a Gate 2 failure: crawlers visit your pages (logs show it) but citations go to competitors whose titles and formats match the question more precisely.

Gate 3: Can the engine extract an answer?

Engines lift chunks of roughly 200-400 words. Your job is to make one chunk self-sufficient:

  • Answer first. State the complete answer in the opening block, then elaborate. A 500-word scene-setting intro pushes your answer below the extraction window.
  • Question-shaped headings. H2s that match how users phrase questions give retrieval a direct hook.
  • Text, not artifacts. Answers that exist only in images, charts without captions, videos without transcripts, or JavaScript-driven accordions don't get extracted.
  • Specifics. Numbers, names, dates, prices. Quotable facts survive synthesis; adjectives don't.

Symptom of a Gate 3 failure: your page appears in the citation list occasionally but your actual points never make it into the answer text, or engines cite you for trivia while your main argument goes unused.

Diagnosing in order

Run the gates as a decision tree. Never mentioned + no crawler visits → Gate 1: fix access. Crawled but not cited → Gate 2: fix the cover and format. Cited but your answer isn't used → Gate 3: restructure the page answer-first. Re-measure after each fix: the tracking method gives you the before/after, or a Lighthouse analysis scores all three layers per page automatically.

Frequently asked questions

How long after fixing robots.txt will AI engines see my content?

Crawlers typically return within days; answer freshness varies by engine. Retrieval-backed engines (Perplexity, ChatGPT with browsing) can reflect changes within days, while training-data knowledge updates only with model releases.

Do accordions and tabs really hurt AI citations?

Content hidden behind JavaScript interactions is frequently invisible to AI crawlers, which mostly don’t execute JS. If the answer only exists inside an accordion, assume engines can’t extract it. Render the text in the initial HTML.

Does schema markup help AI citations?

It helps machine readability: FAQPage schema marks Q&A structure explicitly and article schema carries dates and authorship. It’s low-effort, worth doing, and standard on this site, but it doesn’t rescue content that fails the three gates.

Why does a competitor with worse content get cited more?

Usually a Gate 2 issue: their title, slug, and format match the question’s shape better, or they appear in third-party sources (Reddit, review sites, listicles) that engines weight heavily. Precision and format beat general quality in retrieval.