Shopify agents.md, llms.txt, and robots.txt: What Merchants Actually Need for AI Shopping

The short answer: Shopify already creates the files

Shopify says every store automatically serves three discovery addresses: /agents.md, /llms.txt, and /llms-full.txt. By default, they return the same information, including the store name, store address, sitemap, policy links, and other places an AI tool can use to understand the store.

Shopify calls /agents.md the main source for this store-level information. The two llms files are compatibility versions for older crawlers that look for those names. A merchant can open each address in a browser and inspect what Shopify is already providing.

For most stores, that is enough at the file level. Shopify’s guidance is direct: you do not need a third-party app to create these files. Customizing them should come after you find a specific information gap, not before.

• /agents.md is Shopify’s main store guide for AI agents.
• /llms.txt is a compatibility address for tools that expect that filename.
• /llms-full.txt is another compatibility version that Shopify serves automatically.
• All three show the same content by default.
• A paid generator is not required just to make the files exist.

The three layers do different jobs

The filenames sound more complicated than the jobs they perform. Think of Shopify Catalog as the product file, agents.md as a short guide to the store, and robots.txt as a set of access instructions for open-web crawlers.

The store guide can point an AI tool toward policies, discovery endpoints, and general store information. The access file can tell a named crawler which paths it may visit. Neither one replaces the product titles, descriptions, options, images, prices, and stock details carried through Shopify Catalog.

Keeping those jobs separate prevents a common mistake: editing robots.txt or adding a long llms.txt file when the real problem is a vague product title, a missing size option, an old price, or a return policy that is hard to find.

Shopify Catalog is the product-data layer

Shopify describes Catalog as the primary way its agentic storefronts receive product data. Eligible products can be shared with AI channels using structured details such as title, description, options, images, price, availability, and other attributes.

For ChatGPT specifically, OpenAI says Shopify product data is already integrated through Shopify Catalog and that individual Shopify merchants do not need to provide an additional direct feed. Product results are still selected independently, so integration is not a promise that every item will appear or hold a particular position.

The practical work is to make the product facts worth sharing. A branded title that never names the item, a colour hidden in prose, a default variant label, or a stale stock value can still make the product harder to interpret. Catalog delivery can move data reliably without improving weak source data.

• Use product titles that say what the item is.
• Put useful details such as size, colour, material, fit, or compatibility in clear fields.
• Keep price and stock current for each option when they differ.
• Use images that match the product or selected option.
• Keep seller, shipping, return, privacy, and contact information easy to reach.

Catalog Mapping matters when your product data is custom

Many Shopify stores use custom fields behind the scenes. Shopify calls these metafields and metaobjects. A store might keep a more descriptive product name, a technical specification, or a grouping value in one of those fields instead of the standard product record.

Shopify Catalog Mapping lets the merchant choose which source should supply the catalog title, description, and category. It can also help stores that group related products using custom fields, product tags, or separators inside titles. Shopify says every store can use mapping, but many stores will not need to change the defaults.

Use mapping when you can name the problem it solves. For example, the storefront title may be intentionally short while a verified custom field contains the clearer product name. Or several separately managed items may need a reviewed grouping rule. Preview the result before saving and check a representative set of products rather than assuming one mapping fits the entire catalog.

• Review where Catalog gets the title, description, and category.
• Use verified custom fields only when they contain better product facts.
• Check grouping when products use unusual title patterns, tags, or custom fields.
• Preview several ordinary and unusual products before saving.
• Do not invent missing attributes just to fill a mapping.

robots.txt affects the open web, not Shopify Catalog

Shopify stores can customize robots.txt to guide named web crawlers. This affects whether those crawlers are invited to visit product pages through the open web. Shopify notes that these instructions are advisory and that not every crawler is certain to follow them.

Shopify also draws a firm boundary around this setting. Blocking an AI crawler in robots.txt or at the network layer affects open-web discovery, but it does not stop Shopify Catalog from sending product data to an agentic storefront the merchant has activated.

That means robots.txt is not a master switch for every AI channel. Edit it only when the store has a clear access policy and understands the paths and crawlers involved. Shopify handles the network-level bot layer for stores and does not recommend placing a proxy in front of Shopify as a crawler fix.

• Review the current robots.txt rules before adding new ones.
• Confirm which product paths a rule would affect.
• Separate open-web access from Shopify Catalog channel settings.
• Do not assume blocking one crawler removes products from every external source.
• Test theme-level changes in a duplicate theme and validate before publishing.

What to check before paying for another app

Start with the store you already have. Open the three discovery addresses. Review the Agentic area in Shopify Admin. Preview how ordinary products and unusual products are represented. Then compare the catalog facts with the live product page and policy links.

If the files are present and accurate, leave them alone. If the product data is weak, use Catalog Scanner to inspect an export before changing mappings. If product pages are blocked, unclear, or carrying conflicting page markup, use Site Scanner. If the foundations are viable and the merchant needs Shopify-specific channel help, scope Shopify Agentic Enablement from that evidence.

Customizing agents.md or another theme template can be appropriate when the default store guide is missing important, verified context. Make that change in a duplicate theme, validate the rendered file, and keep an approval checkpoint before it reaches the live store.

• Discovery files missing or incorrect: confirm the theme and Shopify setup before adding an app.
• Catalog fields or grouping are unclear: review Catalog Mapping and scan the source data.
• Product pages or policies are hard to reach: scan the storefront.
• Channel behavior or fit is unclear: review the Shopify pathway before implementation.