Scan Results: https://ai.google.dev/gemini-api/docs

Create an llms.txt file at the root of my documentation site.
Place it at: [domain]/llms.txt

Structure it exactly as follows:
Line 1: # [Product Name]
Line 2: Blank line
Line 3: One paragraph summary of what the product does and who it's for.
Line 4: Blank line
Line 5: ## Technical Constraints
- Rate limits
- Authentication requirements
- SDK language support
- Known limitations
Line 6: Blank line
Line 7: ## Key Documentation
- [URL]: [one-line description] (repeat for top 10 most important pages)

My docs are at: [URL]
My product does: [brief description]

Create an llms-full.txt file at the root of my documentation site.
This should contain the entire documentation merged into a single Markdown file.

My documentation is returning excessive HTML to AI agents.
Implement server-side content negotiation in my app.

When a request includes the header 'Accept: text/markdown',
return ONLY the technical content as clean Markdown.

Strip completely: navigation menus, footers, cookie banners,
sidebar links, breadcrumbs, social sharing buttons,
interactive widgets, and any HTML not containing technical content.

The response should contain only:
- Page title as H1
- Section headings as H2/H3
- Body text
- Code blocks with language tags
- Parameter tables
- No HTML tags, no CSS, no scripts

Reduce layout boilerplate (header, footers, scripts) or use Markdown Content Negotiation to strip DOM noise.

Use structured placeholders like <YOUR_API_KEY> or process.env.

Expose an OpenAPI or Swagger specification at a standard path (e.g. /openapi.json) or link to it using the "Link" HTTP response header with rel="openapi".

Embed a <script type="application/ld+json"> block classifying your docs as a TechArticle, Guide, or APIReference, containing semantically enriched metadata.

Configure and expose an MCP server definition at /.well-known/mcp.json or publish guidelines showing AI agents how to interact with your APIs using MCP tool declarations.

Provide a comprehensive table or section listing error codes, their causes, and solutions. This lets agents cross-reference error stacks and resolve integration errors automatically.