Get cited by ChatGPT, Claude & Perplexity
Last updated: May 15, 2026
TL;DR
AI assistants only cite pages they can fetch, read without running JavaScript, and understand. Allow the AI crawlers in robots.txt, server-render your content, add schema.org JSON-LD and an llms.txt, keep one clear answer per page — then measure it. The five steps below are the whole game.
What “AI visibility” actually means
Classic SEO asks “does Google rank this page?”. AI visibility asks a different question: “when someone asks ChatGPT, Claude, Gemini, Perplexity, Doubao or Grok about your topic, is your page good enough to be the source the model quotes?” The mechanics are different. Assistants reach your site through bots like GPTBot, ClaudeBot, PerplexityBot and Google-Extended, most of which do not execute JavaScript and give up quickly. If the bot cannot get in, cannot see your content, or cannot extract a clean fact, you are invisible to AI — no matter how well you rank on Google.
The five-step checklist
1. Let the AI crawlers in
The single most common reason a site is invisible to AI is the most boring one: it blocks the bots. Check your /robots.txt for a Disallow that catches GPTBot, ClaudeBot, OAI-SearchBot, PerplexityBot or Google-Extended, and check that your CDN or WAF (Cloudflare bot-fight in particular) is not silently returning 403 to those user-agents. Allowing AI training and retrieval on your public pages is what makes you recommendable — see our GPTBot reference for the exact directives, and use Bot Paywall if you instead want to charge specific bots per URL rather than block them outright.
2. Server-render the content that matters
Most AI crawlers fetch raw HTML and do not wait for client-side JavaScript to hydrate. If your headline, body copy and key facts only appear after a framework boots, the bot sees an empty shell. Server-side render (or statically pre-render) the content you want quoted. A fast rule of thumb: view-source your page; if the text a human reads is not in that raw HTML, neither AI assistants nor their crawlers will use it.
3. Add structure: schema, headings, llms.txt
Structure is how a model turns prose into facts it can safely attribute. Add schema.org JSON-LD (Organization and WebSite sitewide; Article, FAQPage or Product per page). Use a single descriptive <h1> and a sequential heading outline. Publish an /llms.txt that points assistants at your best pages, and ping IndexNow so Bing and Copilot pick up changes within minutes instead of weeks.
4. One clear answer per page
Assistants prefer pages that answer one question concretely. Lead with the answer, then elaborate. Name your brand and what it does in the first paragraph — models disambiguate entities from explicit text, not from your logo. Keep enough plain-language text on the page for the model to extract names, numbers and specifics; a page that is mostly imagery or marketing abstraction has nothing for an LLM to cite.
5. Measure it, then keep it fresh
You cannot improve what you do not measure. Audit a URL the way an AI bot sees it — raw fetch, no JavaScript, per-agent — and fix the issues by severity. You can run a free check on the homepage in about 30 seconds, or take the 2-minute readiness quiz if you want a directional score without a URL. Then re-audit after each change: AI visibility is a maintenance habit, not a one-time project.
What this is not
It is not keyword stuffing, it is not a “submit to AI” button, and there is no public API to “register” with ChatGPT or Claude. Anyone selling that is selling snake oil. The only durable lever is making genuinely useful content that is technically reachable and structurally legible. The good news: that is fully in your control, and most sites are losing on the boring technical basics rather than the writing.
FAQ
Does blocking GPTBot hurt my Google ranking?
No — GPTBot is OpenAI's crawler, not Googlebot. Blocking it only removes you from ChatGPT's training and browsing. Your Google ranking is unaffected, but you also lose any chance of being cited inside ChatGPT.
Is llms.txt an official standard?
It is a community proposal (llmstxt.org), not a mandate any vendor enforces. It costs nothing to add and gives assistants a curated map of your best content, so the downside is zero and the upside is real as adoption grows.
How often should I re-audit?
After every meaningful content or infrastructure change, and on a standing monthly cadence otherwise. CDN config, framework upgrades and theme changes silently break AI crawlability more often than people expect.
Want the per-agent breakdown for your own site? Run a free audit or see the plans.