Google ignores llms.txt. Shopify still ships it.

Google Search says it does not use llms.txt for AI Overviews, AI Mode, rankings, or search visibility. Shopify says every Shopify store automatically serves llms.txt and llms-full.txt as compatibility URLs for older AI crawlers that look for those conventions.

Those facts do not conflict. They describe different consumers and different layers of an agentic-commerce stack. Google is explaining what Google Search uses. Shopify is publishing compatibility files while telling merchants that Shopify Catalog remains the authoritative product source for agentic storefronts.

The practical conclusion is narrow: publish or maintain llms.txt when it serves a documented audience, but do not present it as a Google ranking technique or as a replacement for product feeds, structured commerce interfaces, crawlable pages, or checkout tools.

Google’s position is explicit

Google’s guide to AI features and websites says the established foundations of Search remain relevant to AI features. It recommends making content accessible, useful, technically sound, and supported by the appropriate product or business data surfaces.

The same guide states that Google Search does not use llms.txt or special AI markup. Google also says those additions neither help nor hurt rankings or visibility in its AI features.

That wording rules out several common claims:

• llms.txt is not required for Google AI Overviews.
• llms.txt is not required for Google AI Mode.
• Adding the file is not a documented Google ranking factor.
• Removing the file is not a documented Google penalty.
• A crawl of the file by some user agent is not proof that Google Search uses it.

The statement is about Google Search. It should not be generalized into “no agent uses llms.txt” unless evidence supports that broader claim.

What Shopify publishes

Shopify’s agentic storefront product documentation describes several distinct surfaces.

Shopify says /agents.md is the canonical agent-discovery file for a storefront. It also serves /llms.txt and /llms-full.txt as compatibility URLs. By default, the compatibility files contain the same storefront guidance as /agents.md.

Shopify also states that Shopify Catalog is the authoritative source of product data for activated agentic storefront channels. The discovery files are separate from that catalog layer and do not replace it.

That separation is the key to the apparent contradiction. Shopify can publish llms.txt for compatibility even when Google Search ignores it, because Shopify is not claiming that the file controls Google Search.

Five layers that should not be collapsed

Layer	Primary job	Example	What it does not prove
Search discovery	Crawl, index, rank, and present web results	Google Search	That an agent reads llms.txt
Agent guidance	Offer a concise, human-readable map or instructions	agents.md, llms.txt	That product data is complete or transactional
Product data distribution	Supply current product, price, availability, and policy data to activated channels	Shopify Catalog	That a browser agent can modify checkout
Commerce protocol and tools	Expose structured cart or checkout operations	UCP, MCP, WebMCP	That a search engine ranks the site higher
Buyer confirmation and payment	Capture approval, challenges, and final order state	Shopify checkout, Shop Pay, 3-D Secure	That an earlier discovery file authorized a purchase

An implementation can use several layers at once. Their coexistence is normal. Problems start when evidence from one layer is used to claim success in another.

For example, an HTTP 200 response from /llms.txt confirms that the file was served. It does not confirm that an agent fetched it, followed it, found a product, created a cart, completed checkout, or generated revenue.

Why Shopify may keep compatibility URLs

The llmstxt.org v2 proposal, last modified on 10 August 2026, defines a plain-text Markdown file intended to help agents use a website. The project presents the format as a proposal and documents its suggested syntax and discovery behaviour.

Serving a compatibility URL can be reasonable when software, developers, or experimental agents already look for that location. A platform can reduce integration friction by mapping several discovery conventions to the same underlying guidance.

But compatibility is not evidence of universal consumption. The strongest defensible statement is that the file is available for consumers that choose to request it.

This distinction mirrors familiar web patterns. A site may publish RSS even though not every reader uses a feed reader. It may expose an API without claiming every visitor calls it. Availability and usage are different measurements.

Discovery guidance is not a product feed

A good discovery file is compact, stable, and selective. Product catalogs are large, dynamic, and operational. Combining the two creates maintenance and trust problems.

According to Shopify’s documentation, Catalog is the authoritative product-data layer for agentic storefront channels. That is where a merchant should expect platform-managed product synchronization to matter. A storefront discovery file can point to policies, collections, documentation, support resources, or structured interfaces, but it should not be treated as a second, manually maintained inventory database.

Duplicating product claims inside llms.txt creates predictable failure modes:

• price information becomes stale;
• availability differs from the active catalog;
• variants are omitted or simplified;
• regional restrictions disappear;
• promotional terms survive after the promotion ends;
• an agent receives conflicting sources with no authority rule.

If a file includes dynamic commercial facts, the owner needs an automated source of truth and an update contract. Otherwise, linking to authoritative resources is safer than copying them.

The robots.txt distinction

Shopify also explains that blocking or allowing open-web crawlers affects open-web access, while products available through activated agentic channels can still be supplied through Shopify Catalog.

This does not make robots.txt irrelevant. It means that crawler access and channel activation are different controls.

robots.txt communicates crawling preferences to compliant crawlers. llms.txt is a proposed guidance file. Shopify Catalog is a platform-managed product-data surface for activated channels. None is a universal replacement for the others.

Teams should document which control governs which consumer instead of creating one “AI visibility” switch in their mental model.

When publishing llms.txt is reasonable

Publishing the file can be reasonable when all of the following are true:

1. There is a named or testable audience for the file.
2. The content points to authoritative, maintained resources.
3. The file does not promise Google ranking or universal agent use.
4. Its owner and update trigger are defined.
5. Its security review treats file content as untrusted input to downstream agents.
6. Its effect is measured separately from search traffic and sales.

For a Shopify storefront, the platform may already generate compatibility files. Before adding custom automation, inspect the live responses and determine whether Shopify’s generated content satisfies the actual use case.

When it is not the priority

llms.txt is a low-priority task when more fundamental evidence is missing. Examples include:

• important pages are not crawlable or return inconsistent canonical signals;
• product titles, descriptions, availability, or policies are inaccurate;
• Merchant Center or the relevant commerce channel is incomplete;
• structured data contradicts visible content;
• the site lacks clear shipping, return, or support information;
• agent-facing cart and checkout flows have not been tested;
• there is no monitoring for stale links or contradictory claims.

Google’s guidance supports this prioritization: conventional technical and content quality practices remain the foundation for its AI search experiences.

The site’s methodology, validator, examples and creation guide address those separate checks without presenting any of them as proof of Google use.

A measurement plan that separates availability from use

Start with a staged evidence model.

1. Presence

Check that /llms.txt returns the expected status, media type, encoding, and content. Record the observation date and content hash.

2. Validity

Verify syntax, link resolution, language, canonical resources, duplication, and freshness. A valid file can still be unused.

3. Retrieval

Use server logs or another authorized observation source to identify requests. Classify known, unknown, and spoofable user agents carefully. A user-agent string alone is weak identity evidence.

4. Following links

Look for a defensible sequence between a file request and requests to linked resources. Shared IP addresses, caches, and privacy systems can make attribution impossible.

5. Business action

Measure product views, tool calls, cart creation, checkout updates, and completed orders through their own systems. Do not infer these outcomes from the file request.

6. Incremental impact

If the question is whether the file caused an improvement, use a controlled comparison or another explicit causal design. A before-and-after graph without controls may reflect releases, seasonality, channel activation, content changes, or measurement drift.

At each stage, preserve “not measurable” when the available data cannot support a conclusion.

Maintenance checklist

• Identify the intended consumers and document the evidence for them.
• Keep the file concise and link to maintained canonical resources.
• Do not duplicate rapidly changing product, price, stock, or policy data without automation.
• State the language and scope clearly.
• Check every linked URL and redirect.
• Remove claims that imply Google Search uses the file.
• Monitor changes to Google, Shopify, and llmstxt.org documentation.
• Record content hashes and observation dates.
• Treat any text consumed by an agent as potential prompt-injection input.
• Measure retrieval, downstream actions, and business outcomes separately.

What can be claimed today

As of 29 September 2026, the following statements are supported by the cited primary sources:

• Google Search says it does not use llms.txt for its AI features.
• Google says the file neither helps nor hurts Google rankings or visibility.
• Shopify publishes /agents.md as its canonical storefront guidance file and exposes /llms.txt and /llms-full.txt for compatibility.
• Shopify Catalog, not those discovery files, is the authoritative product-data source for activated agentic storefront channels.
• llmstxt.org describes a community proposal for a Markdown guidance file.

The sources do not establish universal agent adoption, a ranking benefit, or a conversion effect.

Conclusion

“Google does not use llms.txt” and “Shopify publishes llms.txt” answer different questions.

Google is defining the inputs to Google Search. Shopify is supporting storefront discovery conventions while routing authoritative product data through Catalog. The llmstxt.org proposal offers a lightweight format that some consumers may choose to use.

The right implementation decision therefore depends on the target consumer. Publish the file for a documented audience, keep it factual and maintainable, and measure it as a discovery surface. Do not market it as a shortcut to Google visibility or as a substitute for commerce data and transactional tools.