What Is llms-full.txt and Do You Need It?

llms-full.txt is one plain-text or Markdown file holding the full text of your key documentation pages, so an artificial intelligence (AI) tool can read them in one request. At Layer3Labs, we publish an llms.txt on this site and on the other sites we build and operate. You should assemble your index file first with our free llms.txt generator, which supplies clean starter files by site type, before you decide whether an llms-full.txt file is worth maintaining.

The convention sits beside the llms.txt proposal, which our guide llms.txt explained covers. A standard llms.txt is a curated table of contents with web addresses (URLs) and brief summaries, while llms-full.txt joins the full text of the listed pages into one large file.

llms-full.txt vs llms.txt

An llms-full.txt file is your whole documentation in one file. An AI model or developer can retrieve your entire library in one Hypertext Transfer Protocol (HTTP) request instead of issuing dozens of sequential fetches across your individual web pages.

A standard llms.txt file points an agent to the right pages with short descriptions and URLs. llms-full.txt holds the page text itself, so the agent does not need to browse further.

The table compares llms.txt and llms-full.txt on five questions:

Who Needs llms-full.txt

You need an llms-full.txt file primarily if you run documentation sites or application programming interface (API) references where users copy and paste the docs into AI tools. In these workflows, users want to paste a single context link into an AI tool instead of copying individual documentation pages.

A small business or marketing site whose pages fit in llms.txt as links gains little from an llms-full.txt file. Markdown versions of each page (page.md) serve the same need page by page, which is Stripe's approach.

Engineering teams that maintain active API docs should consider publishing one, so developers can paste the complete documentation into an AI tool in one step.

Who Publishes One

When we requested llms-full.txt from twelve addresses on October 26, 2026, nine returned a file, all of them documentation sites. The three endpoints that returned a 404 status were Stripe's documentation at docs.stripe.com, the Vercel marketing root at vercel.com, and Layer3Labs at www.layer3labs.io.

The file sizes we recorded run from about 0.47 MB to about 69 MB:

Anthropic at platform.claude.com serves an llms-full.txt file measuring about 47 MB, while its companion llms.txt is only 90.9 KB. Cloudflare at developers.cloudflare.com serves an llms-full.txt measuring about 69 MB. The Vercel documentation at vercel.com/docs provides a file of roughly 11.4 MB. OpenAI at developers.openai.com delivers approximately 9.3 MB. Supabase at supabase.com serves about 7.3 MB. Perplexity at docs.perplexity.ai serves roughly 3.3 MB. Mintlify documentation at www.mintlify.com/docs serves about 1.7 MB. Svelte at svelte.dev serves 1.3 MB, and GitBook documentation at gitbook.com/docs serves approximately 0.47 MB.

Six of the nine working files exceeded 3 MB, making them vastly heavier than their companion indexes. Stripe has one of the largest llms.txt files in our checks at 97.6 KB, yet docs.stripe.com serves no llms-full.txt file at all. Stripe instead offers clean standalone Markdown versions of each individual page at their respective URLs. In addition, Vercel demonstrates that placement can vary: it returns a 404 at the domain root but hosts its 11.4 MB file under its documentation subdirectory.

Where llms-full.txt Came From

The llms-full.txt convention began outside the core specification. Documentation hosting platform Mintlify announced support for /llms.txt and /llms-full.txt on November 20, 2024, noting that it originally developed llms-full.txt in a direct collaboration with Anthropic. The goal was to let programmers paste one link into an assistant instead of manually copying dozens of documentation tabs.

Mintlify and llmstxt.org disagree about whether llms-full.txt is part of the proposal. Mintlify stated in its public announcements that the full-text pattern was included in the official specification. However, the current v2 text published at llmstxt.org makes no mention of llms-full.txt.

During the earlier v1 period of the proposal created by Jeremy Howard at Answer.ai, the project maintained experimental context-expansion tooling called llms_txt2ctx. That tool generated companion files named llms-ctx.txt and llms-ctx-full.txt, which frameworks like FastHTML continue to host. As documented in the llmstxt.org change log, the August 2026 v2 revision removed that context tooling entirely from the proposal. Today, llms-full.txt operates as a widely recognized industry convention rather than a formal specification requirement.

Mintlify says AI agents visit llms-full.txt over twice as much as llms.txt, based on data from Profound.

What Goes Inside the File

An llms-full.txt file contains the complete plain text or Markdown bodies of your documentation pages organized sequentially. Each section carries the full page text along with its title and source URL.

As outlined in the Mintlify llmstxt documentation, each individual page entry appears with its title, canonical source URL, optional description, and complete Markdown content. Vercel structures its full file by separating individual pages with explicit metadata blocks that list the title, description, last-updated timestamp, and source URL. Cloudflare prefaces each distinct page section with structured YAML (YAML Ain't Markup Language) frontmatter. Svelte opens its 1.3 MB full file with a single declarative line identifying the document as the complete developer documentation for both Svelte and SvelteKit.

Because full files can overwhelm context windows, some teams provide tiered options. On svelte.dev, the Svelte team hosts three distinct versions alongside its standard index: llms-full.txt for the comprehensive text, llms-medium.txt labeled as an abridged edition, and llms-small.txt labeled as a compressed edition. Anthropic links its llms-full.txt from the last line of its platform.claude.com/llms.txt index.

How to Create llms-full.txt

Several popular hosting services and content management platforms now generate llms-full.txt automatically alongside your base index. If your documentation runs on a supported vendor, you may already have a live endpoint without needing custom build scripts.

Platforms that provide automatic generation include:

• Mintlify automatically hosts /llms.txt and /llms-full.txt at both the domain root and under /.well-known/. While Mintlify enforces a 100,000-character ceiling on standard llms.txt files, that character cap does not apply to llms-full.txt. Uploading a custom llms-full.txt to your repository root overrides the default output, provided the custom file begins with an H1 heading.
• GitBook publishes an /llms-full.txt file that combines your entire documentation library into one address, automatically incorporating pages that are otherwise hidden from sidebar navigation.
• Shopify serves an llms-full.txt endpoint for every storefront on its platform, displaying identical contents to its standard llms.txt index by default.
• AIOSEO for WordPress includes an llms-full.txt generation toggle, though AIOSEO requires a Pro account to activate the full-text feature.
• llmstxtgenerate.com provides free standard llms.txt generation, but charges a one-time fee of $2.99 to download an exported llms-full.txt file.
• Cloudflare includes automated generation of both llms.txt and llms-full.txt within its AI Index product, which operates in private beta.

Step-by-Step Manual Assembly

If your hosting platform does not build the file for you, assemble llms-full.txt by hand in five steps:

• Start from your llms.txt so the pages match.
• Export each page's clean text or Markdown.
• Join them in one file in the same order, each with an H1 or H2 title and its source URL.
• Save it as llms-full.txt at the root.
• Link it from llms.txt (as Anthropic does).

Common Problems with Full Files

The primary challenge with llms-full.txt is excessive file weight. As shown by Cloudflare's 69 MB payload and Anthropic's 47 MB file, comprehensive documentation is often bigger than an AI tool can read in one go. You can resolve this issue by adopting Svelte's multi-tier pattern, publishing smaller specialized editions, or mimicking Stripe by publishing clean standalone .md endpoints for individual pages.

llms-full.txt also goes stale. Because it copies the full body of your documentation, it falls out of date when an engineer edits an API parameter or an installation guide. Regenerate it every time you publish page changes.

Some generators charge for an llms-full.txt you may not need. AIOSEO needs a Pro account for it, and llmstxtgenerate.com charges $2.99 one-time to download one. Start with a free llms.txt and add llms-full.txt only if you run documentation that people paste into AI tools.

The Practical Decision for Your Site

llms-full.txt is optional. You should build it only after llms.txt exists, and mainly if you run documentation that people paste into AI tools. A small business or marketing site whose pages fit in llms.txt as links gains little from an llms-full.txt file.

Our guidance would change if a major AI tool documents that it reads the file, or if the proposal adds it. Furthermore, Google documentation explicitly states that Google Search itself does not use llms.txt files, and creating them will neither harm nor help visibility or rankings in Google Search. Similarly, the Chrome Lighthouse audit guidance treats missing files as not applicable because adoption remains optional.

If you run documentation that people paste into AI tools, add a build step that regenerates llms-full.txt from the pages in your llms.txt each time you publish.