Why Doesn't ChatGPT Cite My Website?
Published July 23, 2026 · 12 min read
You've published hundreds of blog posts. Your SEO is solid. You rank on page 1 for dozens of keywords. But when you ask ChatGPT, Perplexity, or Claude a question in your niche, your site never shows up as a cited source.
You're not alone. After auditing 1,200+ websites with GeoScore, we found that over 70% of sites are invisible to AI search engines — not because their content is bad, but because they fail one or more technical requirements that AI crawlers need to discover, understand, and cite content.
Here are the 7 most common reasons, ordered from most impactful to least. Fix the first three and you'll likely see your site start appearing in AI-generated answers within weeks.
#1. Your robots.txt blocks AI crawlers
This is the #1 reason by far. When ChatGPT browses the web, it identifies itself as GPTBot. Perplexity uses PerplexityBot. Claude uses ClaudeBot. Google's AI Overviews use Google-Extended.
If your robots.txt doesn't explicitly allow these crawlers, they won't fetch your pages. No fetch = no citation. It's that simple.
How to check:
Open yoursite.com/robots.txt and look for Disallow: / under any of these user-agents. Even a blanket Disallow: / without a specific user-agent block will block everyone, including AI crawlers.
How to fix:
Add these lines to your robots.txt:
User-agent: GPTBot Allow: / User-agent: OAI-SearchBot Allow: / User-agent: ClaudeBot Allow: / User-agent: anthropic-ai Allow: / User-agent: PerplexityBot Allow: / User-agent: Google-Extended Allow: / User-agent: CCBot Allow: /
💡 You can check this in 5 seconds with our free Robots.txt AI Checker.
#2. You don't have an llms.txt file
llms.txt is a markdown file at yoursite.com/llms.txt that gives AI models a curated summary of your site: what it is, who it serves, and which pages matter most. Think of it as an "About page" written specifically for machines.
Without llms.txt, AI crawlers have to guess which pages are important by crawling randomly. With it, you're handing them a prioritized reading list. Sites with llms.txt get cited 2-3x more often in our audit data.
How to fix:
Create a file called llms.txt in your site's root directory:
# Your Site Name > One-sentence description of what your site does and who it serves. ## Core Pages - [Home](https://yoursite.com/): Brief description - [Key Article](https://yoursite.com/article): Brief description - [About](https://yoursite.com/about): Brief description ## Optional - [Blog](https://yoursite.com/blog): Description - [Tools](https://yoursite.com/tools): Description
💡 See our complete llms.txt writing guide or use the llms.txt Checker to verify yours.
#3. Your content isn't structured for AI parsing
AI models don't read web pages the way humans do. They parse the DOM, extract structured data (JSON-LD, schema.org markup), and look for semantic signals. If your content is wrapped in generic <div> tags with no schema markup, the AI has to work harder to understand what your page is about — and it may give up.
The most impactful schemas for AI citations are:
- Article — tells the AI "this is an article, here's the headline, author, and date"
- FAQPage — gives the AI ready-made Q&A pairs it can quote directly
- HowTo — structured step-by-step instructions that AI loves to cite
- Organization — establishes your brand entity and authority
How to fix:
Add JSON-LD to your page <head>. Example for an article:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Your Article Title",
"author": { "@type": "Person", "name": "Author Name" },
"datePublished": "2026-07-23",
"publisher": { "@type": "Organization", "name": "Your Site" }
}
</script> #4. Your content is too thin
AI models prefer to cite content that thoroughly answers a question. If your page has fewer than 300 words of body content, most AI engines will skip it in favor of a competitor's more comprehensive page.
In our audit data, the average cited page has 800-1,200 words of primary content, with clear headings (H2/H3), bullet lists, and direct answers to the question being asked.
How to fix:
- Aim for 600+ words on key landing pages
- Use H2/H3 headings to break up sections (AI parsers use these as section boundaries)
- Put the direct answer in the first paragraph, then elaborate
- Add bullet lists and tables — AI models extract these as "facts"
#5. No E-E-A-T signals
E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) isn't just a Google SEO concept — AI models use similar signals to decide which sources to trust. If your content has no author byline, no "About" page, no privacy policy, and no external citations, AI models treat it as low-trust.
How to fix:
- Add author bylines with links to author bio pages
- Create an
/aboutpage with team info and credentials - Link to your profiles on GitHub, LinkedIn, Twitter (sameAs in schema)
- Cite primary sources and link to .edu/.gov domains where relevant
- Publish a privacy policy and terms of service
#6. You're missing AI-specific discovery endpoints
A new convention is emerging: AI-specific endpoints that give crawlers a machine-readable summary of your site. The two most important are:
/ai/summary.json— a JSON file with your site's name, description, key URLs, and content categories/.well-known/ai.txt— an AI access policy file (similar to robots.txt but specifically for AI training and inference)
Most sites don't have these yet — which means adding them gives you an immediate edge. In our audit data, sites with AI endpoints score 15-20 points higher on average.
How to fix:
Create /ai/summary.json with your site metadata, and /.well-known/ai.txt with your AI access policy. GeoScore's AI Readiness Score tool will tell you exactly what's missing.
#7. Your site is too new or has no inbound links
AI search engines learn about websites primarily through their training data and through real-time web crawling. If your site is less than 6 months old, has no backlinks, and isn't indexed by Google, AI models simply haven't discovered you yet.
How to fix:
- Get indexed on Google first (Google Search Console > Request Indexing)
- Get backlinks from sites that AI models already cite — Reddit, Wikipedia, Stack Overflow, established blogs
- Share your content on platforms AI crawlers monitor: Hacker News, Reddit, Product Hunt
- Link to your site from your GitHub README, social profiles, and directory listings
Note: AI training data updates on a lag. Even after you fix everything, it may take 2-8 weeks for citations to start appearing, depending on the model's refresh cycle.
Find Out Which of These Are Affecting Your Site
Run a free GEO audit on your website and get a 0-100 score across 12 dimensions — plus auto-generated fix files you can deploy today.
Audit My Site Free →No signup required · 12-dimension analysis · Results in 30 seconds
Quick Summary: The AI Citation Checklist
| # | Issue | Impact | Effort |
|---|---|---|---|
| 1 | robots.txt blocks AI crawlers | Critical | 5 min |
| 2 | No llms.txt file | Critical | 15 min |
| 3 | No structured data (JSON-LD) | High | 30 min |
| 4 | Thin content (<300 words) | High | Hours |
| 5 | No E-E-A-T signals | Medium | 1 hour |
| 6 | No AI discovery endpoints | Medium | 20 min |
| 7 | New site / no backlinks | Low | Ongoing |
Have questions? Run a free audit and share your score — we'll help you interpret the results.
Published by GeoScore · Free & open-source under MIT License