How to Get Cited by ChatGPT: a technical checklist

1. Allow GPTBot in robots.txt

Impact: Critical — Without this, ChatGPT cannot access your site at all.

Add these lines to your robots.txt:

User-agent: GPTBot
Allow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

Check yours: yoursite.com/robots.txt

What we measured: 38 of our 48 pages allowed AI crawlers and 5 blocked them. Openness alone gave a 53% citation rate (20 of 38) — necessary, not sufficient.

2. Add Structured Data (JSON-LD)

Impact: High — 18 of the 22 cited pages carried JSON-LD

ChatGPT uses schema.org markup to understand content type and extract key facts.

Minimum viable structured data:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Your Page Title",
  "description": "Clear description of what this page covers",
  "author": {
    "@type": "Organization",
    "name": "Your Company"
  }
}
</script>

What we measured: pages with JSON-LD were cited 62% of the time (18 of 29); pages without were cited 21% of the time (4 of 19). That is a correlation on a small sample, not proof of cause.

3. Content Must Be Readable Without JavaScript

Impact: High — GPTBot doesn't execute JavaScript

If your content is rendered client-side (React, Vue, Angular), GPTBot sees an empty page.

Test: Disable JavaScript in DevTools and reload. Can you read the content?

Solutions:

• Server-side rendering (SSR)
• Static site generation (SSG)
• Progressive enhancement

What we measured: 38 pages open, 5 blocked, 5 undetermined. The open pages were cited 53% of the time (20 of 38).

4. Create llms.txt File

Impact: Medium — Helps but not required

An llms.txt file tells AI crawlers what your site offers. Think of it as a sitemap for AI.

Place it at: yoursite.com/llms.txt

Example:

# YourSite — AI-Readable Site Map

# About Us
> We help developers build better websites.

/blog/
> Technical articles about web development

/tools/
> Free developer tools and calculators

Generate yours: Free llms.txt generator

What we measured: we could not isolate llms.txt as a predictor — 4 of the 48 sites returned a 200 with an SPA fallback page instead of an actual file, which is why the check has to look at the first line, not the status code. Treat llms.txt as a description file, not a citation lever.

5. Page Speed

Not measured in our study. We recorded six fields per page and load time was not one of them, so we have no number to give you here.

The general argument still holds — crawl budgets are finite and slow responses consume them — but treat it as standard web practice rather than something this dataset can back up.

Reasonable targets:

• Compression (gzip/brotli)
• A CDN in front of static assets
• Image dimensions set explicitly
• Fewer redirects

6. Write Content That Can Be Quoted

Word count did not predict citations in our data. Cited pages averaged 1,872 words; never-cited pages averaged 2,122 — the pages that were skipped were slightly longer. Splitting the set at 1,500 words moved the citation rate by 4 points.

What the cited pages did share was quotability: a direct answer near the top, one idea per heading, and claims specific enough to be lifted without context.

• Answer first: state the conclusion before the build-up
• One claim per heading: H2s that read like sentences
• Numbers over adjectives: "6 of 6 runs" beats "highly effective"
• Dates on facts: so a model can tell whether it is stale

This is an observation from 48 pages, not a controlled experiment. We publish the raw records so you can check the claim yourself.