llms.txt

The format and its status

The convention is documented at llmstxt.org: an H1 heading naming the site, a one-paragraph blockquote summary, then markdown sections of curated links, each with a short description written for a machine deciding whether to fetch. It is proposer-led, not standards-body-led, as of mid-2026 — adoption has spread across SaaS and content sites, several AI vendors have acknowledged the file in public notes, and no engine guarantees consumption. The cost of shipping one is a few hundred lines of static text, which is why the upside-versus-cost math favors having it even while the convention matures.

Why it matters for ecommerce

Ecommerce platforms emit sprawling structures by default — product pages, collection pages, tag archives, policy pages. A crawler reading the site cold spends its budget on pages the brand does not especially want surfaced. A curated llms.txt nudges the crawler toward the PDPs and guides the brand wants cited, in the brand's own ranking.

It is also the fastest signal for changes: when a brand rebuilds PDPs or reorganizes collections, updating llms.txt is a one-line change, while waiting for the engine to re-derive the site's hierarchy can take weeks. For Shopify stores — where theme structure is not fully controllable — the file is one of the few crawler-guidance levers entirely in the merchant's hands.

A curated file in practice: an example

A climbing-rope brand's llms.txt opens with a two-sentence brand description, then sections for "Flagship ropes," "By discipline," "Gear care," and "About," each with five to eight curated links. When an LLM crawler retrieves the file it has a ranked reading list: the 9.2mm flagship PDP, the alpine collection, the rope-care guide. That ordering shapes which pages enter the engine's retrieval set — and, weeks later, which URLs get cited when a shopper asks for rope recommendations. The file is regenerated whenever flagship URLs change, as part of the release checklist.

How it relates to neighboring terms

llms.txt serves AI discoverability — helping engines find the right pages — and complements Product schema, which makes those pages parseable once fetched. It works alongside robots.txt, not instead of it: robots.txt grants or denies access per bot, llms.txt suggests priorities to bots already admitted. Crawler admittance itself is checked in the agent-readability score.

How eCommerce Insights uses it

Presence and freshness of llms.txt are checked in every store scan, and the free llms.txt Generator builds a curated file from a store URL — flagship products first, collections, then policies. The platform flags the file when catalog changes make it stale.

Related terms

• AI discoverability — the outcome a curated reading list supports.
• Product schema — the structured counterpart once a page is fetched.
• Agent-readability score — where crawler-guidance checks live.
• GEO (Generative Engine Optimization) — the discipline this file is one tactic within.
• AI visibility — the downstream metric llms.txt contributes to.