llms.txt for agentic commerce: what it is, whether it works, and how to ship one

Published 2026-06-28. Every claim links to a primary source.

Short answer: llms.txt is a small Markdown file at the root of your site (/llms.txt) that hands a language model a curated map of the pages most worth reading. It was proposed by Jeremy Howard of Answer.AI on 3 September 2024 and lives at llmstxt.org. The honest status in 2026: publishing one is now common and cheap, but no major AI vendor has publicly confirmed that it reads your file. For an agentic-commerce site, that makes it forward-compatible insurance — a stable, machine-readable contract you control — rather than a guaranteed traffic channel. Here is what it is, who reads it, and how to ship a good one. For the deeper reference, llmtxt.info maintains a neutral, sourced guide to what llms.txt is.

What llms.txt actually is

Howard’s argument was simple: a language model has a small context window compared with a full website, so asking it to crawl everything to answer one question is wasteful and lossy. Sites already publish robots.txt for crawlers and sitemap.xml for search engines; llms.txt proposes a parallel convention aimed specifically at LLMs. The format is deliberately minimal: an H1 with the site name, a one-line summary in a blockquote, then Markdown sections of curated links with short descriptions. It is meant for inference — the moment a user asks an assistant about you — not for training, and it is curated for signal, not coverage (Answer.AI, llmstxt.org).

Here is a real one — the file this very site publishes at /llms.txt, trimmed for length:

# Agentic Commerce Protocol Info

> Neutral, up-to-date reference comparing the open standards
> behind agentic commerce: ACP, UCP, AP2, MCP and x402.

## Standards
- [Agentic Commerce Protocol (ACP)](/standards/acp): OpenAI + Stripe checkout.
- [Universal Commerce Protocol (UCP)](/standards/ucp): Google + Shopify.
- [x402](/standards/x402): Coinbase HTTP-native stablecoin payments.

## Core pages
- [ACP vs UCP vs AP2 vs MCP vs x402](/compare/acp-vs-ap2-vs-mcp): comparison table.
- [How AI agents pay](/how-ai-agents-pay): the agentic payments stack.

Figure 1 — A live llms.txt: H1, a blockquote summary, then curated sections of links. See the full file at /llms.txt.

How it differs from robots.txt and sitemap.xml

The three files are complementary, not interchangeable. The clearest framing: robots.txt says what a crawler may fetch, sitemap.xml says what exists, and llms.txt says what is worth reading. Only the first two are formal standards. llmtxt.info keeps a fuller breakdown of llms.txt vs robots.txt vs sitemap.xml.

File	Audience	Purpose	A standard?
`robots.txt`	Web crawlers	Access control (allow / disallow)	Yes — IETF RFC 9309
`sitemap.xml`	Search engines	Discovery — list every page	Yes — sitemaps.org
`llms.txt`	LLM assistants	Curation — what to read first	No — community proposal
`llms-full.txt`	LLM assistants	Inlining — the actual content in one file	No — community proposal

Sources: IETF RFC 9309, llmstxt.org.

Does it actually work? The honest, two-sided answer

This is where most write-ups oversell. The supply side and the demand side are moving at very different speeds, and conflating them is the single most common mistake.

Figure 2 — Publishing llms.txt is common; vendors consuming third-party files is not documented. The ~10% adoption figure is from a 300,000-domain analysis reported by Search Engine Journal (Nov 2025). llmtxt.info keeps the sourced answer to whether Google and ChatGPT use it.

So why ship one at all? Because the cost is minutes and the downside is zero. Documentation platforms already auto-generate it; agent frameworks and RAG pipelines increasingly look for it; and when consumption does arrive, a curated, accurate file is a contract you already control. For a commerce brand that also cares about being correctly read by shopping agents — the subject of our companion piece on how AI agents find and read your products — it is a low-effort complement to schema.org and clean server-rendered data.

Watch: llms.txt vs robots.txt vs sitemap.xml

Webflow’s short explainer is a clear walk-through of what the file is, how it differs from the two files it’s often confused with, and how to publish one.

Source: Webflow — “What Is an llms.txt file?” (official channel).

How to ship a good one for a commerce site

The whole exercise is curation. List the 10–20 pages an assistant would need to answer a real buyer’s question, write one clean line each, and leave everything else out.

Start with identity — an H1 with your brand and a one-sentence blockquote of what you sell and to whom.
Group the essentials — collections and flagship products, shipping & returns, sizing/spec guides, and key policy pages, each as a link with a short note.
Curate, don’t dump — omit cart, checkout, login, search and thin pages. Signal beats coverage.
Link canonical HTML (or clean .md versions), keep it UTF-8 and lightweight, and update it when important content changes.
Consider llms-full.txt only if your catalogue copy is small enough to inline usefully.
Validate before you ship so a parser reads it the way you intend.

llmtxt.info publishes a step-by-step guide with copy-paste templates, and you can validate your file against the spec before deploying it.

Where it fits with the agentic-commerce stack

llms.txt is the legibility layer; the protocols on this site are the transaction layer. Pair a clean file with the rest: read how agents find and read your products, compare the payment standards, and see how to accept agent payments. The dated timeline is on the changelog.

Frequently asked questions

What is llms.txt?

A Markdown file placed at the root of a site (/llms.txt) that gives large language models a short, curated map of the pages most worth reading. It was proposed by Jeremy Howard of Answer.AI on 3 September 2024 and is documented at llmstxt.org. It is a community convention, not an IETF or W3C standard.

Does Google or ChatGPT actually read my llms.txt?

As of mid-2026, no major LLM provider (OpenAI, Anthropic, Google, Perplexity) publicly documents consuming third-party llms.txt in its crawler or inference pipeline, and Google’s John Mueller has publicly questioned its value as a search signal. Adoption on the publishing side is real and rising; consumption on the vendor side is unconfirmed. Treat it as cheap, forward-compatible insurance, not a guaranteed traffic channel.

Is llms.txt the same as robots.txt?

No. robots.txt is an access-control standard (IETF RFC 9309) that tells crawlers what they may fetch; sitemap.xml lists what exists for search engines; llms.txt curates what is worth reading for an LLM. They are complementary — keep all three.

What should a commerce site put in llms.txt?

A one-line description, then curated links to the pages an assistant needs to answer buyer questions: product and collection overviews, shipping and returns, sizing or spec guides, integration or standards pages, and key policy pages. Curate for signal, not coverage — leave out cart, checkout, login and thin pages.

Stay current on the standards

A note when a standard moves (ACP, UCP, AP2, MCP, x402). No spam.