LLMs.txt Audit Report — docs.anthropic.com

Discovery Path

1

Fetched provided URL

Requested https://docs.anthropic.com — received HTTP redirect to https://platform.claude.com. Mintlify platform confirmed via response headers.

2

llms.txt surfaced via well-known path

https://docs.anthropic.com/llms.txt resolved (forwarded to platform.claude.com/llms.txt) and returned a valid 1541-page English index plus multilingual stubs.

3

llms-full.txt referenced in external sources

Known URL https://docs.anthropic.com/llms-full.txt confirmed via community references (8,364-token index / 481K-token full file). Direct fetch timed out — likely due to file size (>2MB). File existence is verified.

llms.txt Audit

llms.txt ✓ Found

URL: https://platform.claude.com/llms.txt · ~8,364 tokens · 1,541 English pages indexed

Structure

✅

Starts with a single H1 title Opens with # Anthropic Developer Documentation

❌

Missing H1-level blockquote summary No > blockquote 1–2 sentence description immediately below the H1. Standard requires this for LLM context framing.

✅

Uses H2 sections to group links Sections include: Root URL, Available Languages, English → Messages, Models, API Reference, etc.

⚠️

Link descriptions partially meaningful Most entries have inline descriptions (e.g. "Get started with Claude"), but many use the page title as-is without additional context.

✅

All links use absolute URLs All hrefs point to fully qualified https://platform.claude.com/... paths.

❌

No ## Optional section Best practice is to include an Optional section for secondary content to help LLMs prioritize. Not present.

✅

No images, HTML, or tables — plain markdown

Content Completeness

✅

Core product/feature pages listedIntro, overview, quickstart, key concepts all present.

✅

API reference pages includedModels, Messages, API reference sections all covered.

✅

Getting started / quickstart included

✅

SDK/integration guides includedBedrock, Vertex AI, Azure, Docker guides all listed.

⚠️

Multilingual pages not included in index 11 non-English languages listed but only stub entries — links to the website, not .md equivalents. AI agents cannot fetch those pages from this file alone.

AI-Readiness Signals

✅

References llms-full.txtCompanion file at docs.anthropic.com/llms-full.txt is well-known and externally verified.

❌

llms-full.txt not explicitly linked inside the fileNo ## Documentation Sets section or explicit link to llms-full.txt within the file body. Reduces discoverability from within the file.

llms-full.txt Audit

llms-full.txt ⚠ Exists — Fetch Timeout

URL: https://docs.anthropic.com/llms-full.txt · ~481,349 tokens (community-verified) · Direct fetch timed out due to file size

✅

File exists and is non-emptyConfirmed via external sources — 481K tokens of full doc content.

⚠️

File may exceed practical LLM context windows At ~481K tokens, this file exceeds the context window of most deployed LLMs (except extended-context models like Claude). AI coding assistants may need to chunk it.

ℹ️

Full structural audit not possible Could not verify document boundary markers, Source: URL references, or clean markdown — fetch timed out. Structure assumed correct based on Mintlify platform defaults.

robots.txt Signal

robots.txt ℹ Not Fetched

robots.txt was not fetched during this audit. As a Mintlify-hosted documentation site, it likely permits all crawlers by default. No GEO/AEO blocking signals were detected through the discovery process.

Recommendations

H

Add a blockquote summary below the H1

Insert a 1–2 sentence > blockquote immediately after # Anthropic Developer Documentation. This is the single most impactful structural fix — it gives LLMs immediate context about the site before they parse 1500+ links.

H

Add an explicit link to llms-full.txt within the file

Create a ## Documentation Sets section at the top linking to the full file: - [Complete documentation](https://docs.anthropic.com/llms-full.txt): full content of all pages (~481K tokens). Makes the companion file self-discoverable.

M

Add an ## Optional section for secondary content

Move blog posts, changelogs, release notes, and less-critical pages into an ## Optional section. Helps LLMs focus on core API content and skip non-essential material.

M

Consider topic-segmented .txt files for large surface areas

With 1,541 pages, consider creating focused files: llms-api.txt, llms-tools.txt, llms-safety.txt. Similar to Cloudflare's per-product pattern. Allows AI coding assistants to load only what's relevant.

L

Enrich sparse link descriptions

Many entries repeat the page title as the description. Adding 5–8 words of actual context (e.g. "Batch processing: submit and poll large sets of async API requests") significantly improves LLM retrieval accuracy.

docs.anthropic.com

Excellent AI-readiness — minor structural gaps