A vendor says llms.txt improves AI visibility. What should you ask?

You may have arrived after an AI audit marked llms.txt as missing, or after a vendor offered to add the file for better visibility. Shopify already publishes discovery files for its stores, so the first question is whether the offer names a consumer and evidence that can be checked.

In May 2026, Ahrefs checked whether anything requests it. The study covers 137,210 domains using Ahrefs Web Analytics, which means server log data rather than a survey or a crawl.

Ask for evidence before buying the work

Ask the vendor or the theme developer responsible for the proposal:

1. Which named AI system is expected to request the file?
2. What server-log query will show requests to /llms.txt from that system on this store?
3. What decision changes if the log remains empty?

An answer that only says the file is an AI-readiness requirement is insufficient. Take no action when the offer only recreates Shopify’s default files or promises general shopping visibility. A custom theme file is relevant only when a named consumer or storefront requirement needs different content and the owner will maintain it like code.

What the logs show

Of the roughly 38,000 domains in the sample that published a valid llms.txt file, 97% saw no requests for it whatsoever during May 2026. About 1,100 domains, three percent of those with a file, received any traffic to it at all.

Among the files that did receive traffic, 96% of requests came from bots and 4% from humans.

The composition of that bot traffic is the part that settles the question. AI bots across all four categories that Ahrefs tracks, meaning agents and infrastructure, training crawlers, assistants and retrieval bots, accounted for 19.5% of requests. Broken out, AI agents and infrastructure were 10.5%, training crawlers 5.3%, and AI retrieval bots, the category tied to answering a live search query, were 1.1%.

The largest single source was SEO audit tools at 21.7%. General crawlers took 13.1% and technology profiling tools 11.6%.

So the file is mostly unrequested, and where it is requested, the requester is more likely to be an SEO tool checking whether the file exists than an AI system reading it.

The study’s author recommends against creating llms.txt for AI search visibility, while allowing that this could change if agents come to mediate search results differently. The sample also skews toward technical and SEO-aware sites, and a request is only evidence that the file was fetched. It does not prove that a bot used or acted on the content.

Where the file fits in Shopify commerce

For a store specifically, the finding matches how the commerce surfaces are built.

Shopify’s Global Catalog MCP lets agents search products across Shopify merchants through structured catalog tools. For ChatGPT specifically, Shopify states that eligible merchant products are made available through Shopify Catalog.

Shopify also provides /agents.md, /llms.txt, and /llms-full.txt for storefront discovery. Its product discovery documentation says those files help AI agents understand a store, while Shopify Catalog remains the authoritative product-data feed.

OpenAI separately documents a structured product feed for supported file-upload integrations. That specification does not turn direct feed submission into a task for every Shopify merchant.

The server-log study therefore answers a narrower question. It gives no evidence that a merchant-authored llms.txt file improves AI shopping visibility. It does not prove that no discovery system can consume the file.

What Shopify already supplies

Since May 28, 2026, Shopify storefronts include a default agents.md file. The paths /llms.txt and /llms-full.txt point to that content unless a theme supplies a more specific template, according to Shopify’s developer changelog.

A Shopify merchant therefore does not need an app to create the default files. Customizing them is theme work, and the server-log evidence above gives no basis for treating that customization as a shopping-visibility result.

What the same effort buys elsewhere

Two things in the same time budget have a documented consumption path.

robots.txt directives are read. Crawler access is a real gate, and it is one merchants often close without noticing. Anthropic documents ClaudeBot for training, Claude-User for user triggered fetches and Claude-SearchBot for search indexing, all of which respect robots.txt. A blanket rule, an inherited default from an edge provider, or a line added for an unrelated reason will keep a named crawler out, and the resulting absence looks like nothing at all from inside the admin.

Structured product fields are read. Keep titles, descriptions, identifiers, variant options, prices, availability, and images accurate in Shopify. Shopify Catalog is the documented merchant path to ChatGPT, so the source catalog is the place the merchant can improve.

The general form of the check

llms.txt is one instance of a pattern worth applying to any AI visibility tactic before spending time on it.

Ask which system reads this, by what route, and what evidence exists that the route is used. When the answer is a proposal that no documented consumer implements, the tactic has no mechanism, and no amount of care in the file changes that.

The observable result here is a bounded choice: keep Shopify’s default, or approve one maintained customization for a named consumer and log check. The Ahrefs study supplies a base rate for requests across its sample. It does not diagnose this store or prove ranking, traffic, or sales.