Industry

Shopify Ships llms.txt and Agentic Discovery Files to Every Store

In early May 2026, Shopify shipped three new routes to every store on the platform, with no public announcement, no documentation, and no merchant opt-in required.

The three files are now live at the root of every Shopify store:

  • /llms.txt: links to the store’s product catalogue (/collections/all), search endpoint, agent instructions, and MCP/UCP commerce hooks for agent-driven checkout
  • /agents.md: a Markdown companion to llms.txt
  • /sitemap_agentic_discovery.xml: an XML index referencing llms.txt and other AI-facing pages

The rollout is customisable: merchants can override the default by creating a templates/llms.txt.liquid file in their theme, using the same pattern as robots.txt.liquid. Merchants who had third-party redirect or proxy app workarounds may find them silently overwritten.

What the files contain

The boilerplate is largely identical across all stores. Store name, contact email, and phone number (pulled from store settings) are the only fields that vary between merchants. The capability claims, endpoint structure, and agent instructions are standardised.

That includes advertising UCP and MCP commerce hooks regardless of whether the merchant has agentic storefronts switched on. A store with no AI channel integrations gets the same capability declarations as one actively selling through ChatGPT.

Contact details in the default file are drawn directly from store settings, which may include internal or support-facing email addresses rather than customer-facing ones. Merchants should audit what is being served before assuming defaults are appropriate.

The agentic storefronts dashboard

Alongside the file rollout, Shopify launched an agentic storefronts dashboard in the admin showing sessions and sales attributed to ChatGPT, Perplexity, and Microsoft Copilot over a 30-day window. It includes a listing completeness assessment covering product descriptions, taxonomy, reviews, images, and policies.

Shopify reports AI-driven traffic to its stores has grown eight times year over year, with orders from AI-powered searches up fifteen times.

The dashboard reflects Shopify’s broader Agentic Storefronts product, introduced in the Winter ‘26 edition and made available to all stores by late March 2026. The feature syndicates product catalogues to AI shopping surfaces without merchants needing to build bespoke integrations.

Crawler behaviour

Research from WISLR, tracking bot activity across Shopify stores over a 60-day window, found that major AI crawlers are not reading any of the new agent-discovery paths.

BotActivity
BingbotLLM93 fetches (legacy paths only)
GPTBotLLM71 fetches (legacy paths only)
ClaudeBotLLM28 fetches (legacy paths only)
Microsoft-Commerce-PlatformOnly bot polling /.well-known/ucp

All three major crawlers repeatedly hit /pages/llms-txt and /a/llms, both legacy paths from third-party app workarounds predating Shopify’s native implementation. None fetched /llms-full.txt, /agents.md, or /sitemap_agentic_discovery.xml.

The exception is Microsoft’s commerce platform, which polls /.well-known/ucp on a roughly weekly cadence from Azure infrastructure. This reflects Copilot’s UCP integration with Shopify’s checkout system: an actual transactional connection, not a discovery crawl.

What this means for merchants

The infrastructure is live, but the retrieval signals are not. Merchants should not expect the new files to improve AI visibility in search or AI-generated answers; the crawlers driving citation-based traffic are not reading them.

Catalogue quality is where the actual signal lives. Shopify’s listing completeness framework (descriptions, taxonomy, structured data, reviews, images, return policies) maps directly to the factors that influence whether products appear in AI-generated shopping results. That work has independent value regardless of the new files.

For merchants who want control over what the default llms.txt advertises, the templates/llms.txt.liquid override is available now. There is no urgency for most stores, but checking that contact details in the default file are customer-appropriate is a reasonable precaution.

Sources

More news