SEMPITE Research · Ecosystem Snapshot · August 2026

98% of Shopify Stores Serve the Same llms.txt. It’s Not a Content Map.

llms.txt was designed as a curated map — a short file where a site points AI at its most useful pages. Then the biggest ecommerce platform on earth turned it into something else. We fetched the llms.txt of 10,858 Shopify stores. The file most of them serve is byte-for-byte the same, injected automatically, and it doesn’t map content at all — it tells an AI how to check out. Most merchants don’t know it’s there.

98%of canonical Shopify stores serve the identical auto-file
6,016stores serving the byte-same “Agent Instructions” manifest
1.5%that publish a genuine curated content map
89.7%of Shopify llms.txt files are auto-generated boilerplate

What 7,133 Shopify llms.txt files actually contain

Of 10,858 stores we checked, 7,133 returned a file. Here is what those files are:

Shopify auto-file (“Agent Instructions” + “Store Unavailable”)89.7%
HTML page (not a text file)3.9%
Misplaced robots.txt content2.5%
Other / empty2.4%
Genuine curated content map1.5%

Only 104 of 7,133 stores publish the kind of file llms.txt was created for. The other 98.5% are boilerplate, an HTML fallback, or a misfiled robots directive. To confirm this is a platform default and not a quirk of our sample, we ran a separate check against 1,907 canonical *.myshopify.com stores: 98% served the identical “Agent Instructions” file. Shopify is generating this for essentially every store.

The file isn’t a map. It’s a checkout API in prose.

Every one of those 6,016 files opens the same way — “Agent Instructions — how AI agents can interact with [brand]’s online store” — and then describes Shopify’s Universal Commerce Protocol and the shop.app checkout skill. It is written for an AI that buys, not one that reads:

# Agent Instructions — [Store Name] ## Commerce Protocol (UCP) This store implements the Universal Commerce Protocol (ucp.dev)… ### Typical Agent Flow 3. Cart — use `create_cart` to add items 4. Checkout — use `create_checkout` to start the purchase 6. Complete — use `complete_checkout` (buyer must approve payment)

That is a genuinely forward-looking idea — agentic commerce is coming, and a machine-readable checkout is a reasonable thing to publish. But it is not what the filename says it is. A tool built to read /llms.txt expecting a content index gets a payment protocol instead. And because Shopify serves it automatically, the merchant never made the choice, never curated a single link, and in most cases has no idea the file exists.

Why this matters beyond Shopify

The web’s public directories of “sites with llms.txt” are full of hand-crafted documentation maps — the visible, early-adopter tip. Underneath it, quietly, the single largest bloc of llms.txt files in existence is now an auto-injected commerce manifest that no human wrote. The format is being defined less by its spec than by whoever ships it at scale by default — and right now that’s Shopify.

For anyone selling on Shopify: your llms.txt is already decided for you, and it’s a checkout file — not a description of who you are, what you’re known for, or why an assistant should recommend you over a competitor. As our citation study showed, that manifest does nothing to make an AI mention your brand; being recommendable is a separate job entirely. The auto-file handles the sale after the recommendation. It does not win you the recommendation.

Not sure what your llms.txt says — or whether AI describes your brand the way you’d want?

Run a free AI visibility check

Methodology: on 5 August 2026 we fetched /llms.txt for 10,858 Shopify stores — 8,951 custom-domain storefronts compiled from two public Shopify store lists (skewed toward .in and .com; region and source noted for transparency) plus a cross-check sample of 1,907 canonical *.myshopify.com stores. Files were classified by status, content type and body text; the two Shopify auto-templates were matched on their fixed titles and Universal Commerce Protocol markers (ucp.dev, create_checkout, shop.app), and the “content map” bucket was re-fetched and hand-categorized to separate genuine curated maps from “Store Unavailable” auto-files and misplaced robots directives. Percentages are of the 7,133 stores that returned a file. Single snapshot; a store list is a sample, not a census, and public lists over-represent active mid-market merchants. Dataset released under CC BY 4.0 — cite as “SEMPITE Shopify llms.txt Snapshot, August 2026.”