Your Website Is Invisible to ChatGPT. Here's How to Fix That.

Ask ChatGPT what your company does. Go ahead, try it right now.

If the answer is vague, wrong, or "I don't have information about that" -- you have a problem that no amount of SEO will fix. Traditional search optimization gets you ranked on Google. It does nothing for whether an AI can accurately describe your business when someone asks.

And people are asking. ChatGPT alone has over 400 million weekly active users. Google's AI Overviews show up on billions of searches. When someone types "who builds custom SaaS platforms?" into Claude or Perplexity, the AI either knows you or it doesn't. There is no page 2 to scroll to.

What "LLM-ready" actually means

SEO is about ranking. LLM readiness is about comprehension. AI systems don't rank your page -- they read it, pull out facts, and synthesize answers. Will the AI describe your business accurately? That's the question.

Getting there requires some specific technical work.

1. robots.txt -- let the bots in

AI companies use specific User-Agent strings to crawl the web: GPTBot (OpenAI), ClaudeBot (Anthropic), Google-Extended (Gemini), PerplexityBot. Most websites either have no policy for these, or block them by accident.

What a lot of people miss: OpenAI runs two separate bots. GPTBot collects content for training future models. OAI-SearchBot indexes content for ChatGPT's search results. You can allow one and block the other. Anthropic does the same thing with ClaudeBot and Claude-SearchBot.

If your robots.txt doesn't mention AI bots at all, most will crawl freely. But if you have a blanket Disallow: / for unknown user agents, you might be blocking all of them without knowing it.

2. llms.txt -- a README for AI

This is the newest piece. In September 2024, Jeremy Howard at Answer.AI proposed llms.txt -- a markdown file at your site root that tells AI systems what your site is about and where to find things. Company name as an H1, a short summary in a blockquote, then sections linking to your pages with one-line descriptions.

Why markdown? It's 80% more token-efficient than HTML. The AI gets structured context about your business without wading through navigation menus, JavaScript bundles, and cookie consent dialogs.

There's also an optional llms-full.txt for the full documentation dump. Platforms like Mintlify, Fern, and ReadMe already generate both files automatically.

3. Structured data in JSON-LD

Schema.org markup gives AI systems machine-readable facts about your business. The format matters here: JSON-LD lives in a <script> tag in the raw HTML source. Microdata and RDFa are embedded in rendered HTML. The problem?

AI bots do not run JavaScript.

An analysis of over 500 million GPTBot requests found zero evidence of JavaScript execution. The bot grabs raw HTML and leaves. If your structured data gets injected by client-side JS, it doesn't exist as far as any AI crawler is concerned.

Start with Organization schema (name, URL, description, logo) and WebSite schema. Publishing articles? Add Article schema. Selling products? Product schema with offers.

4. Server-side rendering

Same JavaScript problem, bigger consequences. If your site is a single-page app that renders everything client-side, AI crawlers see <div id="root"></div> and a JavaScript bundle they won't touch. No text. No metadata. Nothing.

Server-side rendering or static generation fixes this. When a crawler requests your page, it gets fully rendered HTML on the first response. This is also why frameworks like Next.js, Nuxt, and Astro have been gaining ground -- they do SSR by default.

5. Meta tags as filters

This one surprised me when I dug into the research. LLMs don't just use meta descriptions for display -- they use them as a filter. The process goes: search, read titles and snippets, decide which pages are worth fetching, then pull full content only from the ones that passed.

A vague title or missing description means your page gets filtered out before the AI reads a single word of your actual content. Write titles that say what the page is (30-60 characters). Write descriptions that summarize what's on it (120-160 characters). Include OpenGraph tags.

6. Semantic HTML

One H1 per page. Logical heading hierarchy. Semantic elements like <article>, <section>, <nav>, <main>. Descriptive alt text on images. Meaningful link text instead of "click here."

None of this is new advice. But AI systems lean on document structure more heavily than traditional search engines do, because they're trying to understand relationships between concepts, not just match keywords.

Where do you stand?

I built a free checker that tests all of this -- robots.txt, llms.txt, structured data, meta tags, semantic HTML -- and also asks three LLMs (GPT-4o, Claude, Gemini) what they actually know about your brand right now. Takes about 30 seconds.

I put the whole thing together in one afternoon using Claude Code, which was honestly a fun way to spend a Monday. The report is shareable, and if your score comes back rough, we fix these things regularly.

Check your score here.