agents.md vs llms.txt: what's the difference and do you need both?

If you've spent any time reading about AI shopping agents, you've probably run into two file names that sound almost interchangeable: agents.md and llms.txt. Merchants reasonably ask whether they need one, both, or whether it even matters. Short answer: they solve overlapping but distinct problems, and as of May 2026, Shopify serves both automatically — but the file being present isn't the same as it being useful.

Where each convention came from

llms.txt emerged first, proposed in late 2024 as a simple idea borrowed directly from robots.txt: a plain-Markdown file at your site's root that gives a large language model a curated, human-written summary of what's on the site and links to the most important pages. It was designed for the general case — any site, not specifically commerce — as a way to help an LLM get oriented without having to crawl and infer everything from raw HTML.

agents.md came out of the more recent push toward agentic commerce specifically — AI agents that don't just summarize content but take actions: browsing a catalog, checking availability, initiating a purchase. It's more structured and commerce-aware by convention, oriented around helping an agent understand a store's products, categories, and policies well enough to act on them, not just describe them.

In practice, think of llms.txt as the general-purpose “here's what this site is about” briefing, and agents.md as the commerce-specific “here's how to actually transact here” briefing. They're complementary, not competing.

What each file actually contains

A typical llms.txt is short: a site title, a one-line description, and a linked list of key sections or pages — think homepage, about, main category pages, contact/support. It's meant to be skimmed in seconds.

A typical agents.md for a Shopify store goes further: store name and description, how the catalog is organized (categories, collections), where policy content lives (shipping, returns, privacy), and enough structure that an agent doesn't have to guess at your store's shape from scratch. Generated well, it should reflect your live catalog — not a hand-written summary that goes stale the moment you add new products.

How Shopify now auto-serves both

As of May 2026, Shopify ships /agents.md and /llms.txt automatically on every store, with agents.md treated as the canonical, commerce-specific file going forward. If you haven't customized anything, your store already has both at those root paths, generated from whatever Shopify's default template produces for your store type. That's a real, useful baseline — it used to be something merchants had to pay third-party apps to generate, and now it's free and automatic.

When you'd want to customize yours

The auto-generated default is generic by necessity — it doesn't know which categories you actually want agents to prioritize, whether you have a distinctive return policy worth highlighting, or which products you'd rather an agent not surface (discontinued lines, internal SKUs, etc.). Customize when:

• Your catalog has non-obvious structure an agent might misread (bundles, made-to-order items, size-run complexity).
• Your policies have merchant-specific terms worth calling out explicitly rather than letting an agent infer them.
• You've made a recent catalog change and want to confirm the file reflects it rather than a stale auto-generated snapshot.

How to check yours

The fastest way to see what your store is currently telling AI agents is to look at the files directly — yourstore.com/agents.md and yourstore.com/llms.txt — and read them the way an agent would: cold, with no context. Do they describe your current catalog? Do they mention policies accurately? If you want a faster gut-check across more than just these two files (structured data, robots.txt rules, feed quality), run a free scan and get a plain-English breakdown of what's solid and what needs attention.

Why the file is only half the story

Here's the thing neither file tells you: whether anything is actually reading them. A perfectly written agents.md that no agent ever fetches is a tree falling in an empty forest. The other half of this problem is visibility — knowing whether GPTBot, ClaudeBot, PerplexityBot, or Google's AI crawlers are actually visiting your store, how often, and whether that traffic is converting into orders. Shopify solved file generation for you. Proof that agents are actually showing up — and buying — is the part still worth measuring yourself.