In early May 2026, Shopify pushed AI discovery files to every active storefront on the platform. Every Shopify store now has a live /agents.md file at its root, with /llms.txt and /llms-full.txt pointing to that same content by default. These files are designed to help AI systems, shopping assistants, and agentic tools understand how to interact with a store. The problem is that Shopify authored them for every store at once. The content is identical across millions of storefronts. You can follow these links to see how the files are identical across a variety of brands from Kylie Cosmetics to Steve Madden to Alo Yoga. This is essentially an uncustomized Shopify template.
What Are llms.txt, agents.md, and llms-full.txt?
These three files serve a common purpose: giving AI systems a structured, machine-readable summary of a website without requiring them to crawl and interpret every page. Each file has a distinct scope.
llms.txt is a markdown file placed at the root of a domain. The llms.txt convention was proposed as a way for site owners to give language models a clean, token-efficient map of the site's most important content: a brief description of who the site is for, links to key pages, and the context an AI needs to accurately understand and represent what the site offers. For an e-commerce brand, a well-written llms.txt would include a factual description of the company, links to priority product and category pages, and enough context to answer "what is this brand and what does it sell?" without the AI having to infer it from product listings. Read more details about how to write your llms.txt in our article here.
llms-full.txt extends that concept with more complete content. Where llms.txt is a concise map, llms-full.txt is intended for AI systems that need deeper context: fuller product descriptions, richer brand context, and more detailed guidance on what the site contains and how it is structured.
agents.md goes a step further. Where llms.txt maps content, agents.md gives behavioral guidance for AI agents that can take actions, not just retrieve information. For an e-commerce store, a well-written agents.md would explain product hierarchies, buying guidance, return and shipping policies, and how an agent should handle common purchase-related queries. It is less a content index and more an operational brief.
Together, these files are meant to function as the brand's AI-facing brief: accurate, specific, and written to help any AI system understand what the brand is, what it sells, and how the agent should interact with this brand’s site on behalf of the user.
What Shopify has shipped is structurally different from the intention of these files. As confirmed in Shopify's developer changelog, the default setup sends identical content across all three paths. The file is platform-generated, not brand-generated. It describes generic store functionality: browse links, search, contact information, and checkout flows. Every Shopify store's default file ends with a line promoting Shopify itself.
The default Shopify AI discovery files contain nothing about which products are your hero products, what your brand stands for, or why an AI system should recommend your store over a competitor.
Why Shopify's Default Files Are Useful but Not Enough
Shopify did something valuable by generating these files automatically. The technical barrier to AI discovery is now zero for any merchant on the platform. Most brands had not created an llms.txt at all, and most were unlikely to, without a native path to do so.
While these default files solve for presence, they do not not solve for differentiation. When AI engines retrieve and rank sources in real time, they evaluate specificity, extractability, and authority. A file that routes AI systems to a generic browse page produces a different signal than one that names the hero product, explains what it does, and links to the page that proves it.
When AI systems compare competing products, they pull from whatever brand signals are available. If two competing brands in the same category both have generic Shopify-generated files with no product-level differentiation, the AI system falls back to whichever has clearer on-page content, stronger structured data, or more authoritative external signals. The discovery file adds nothing to that comparison.
To summarize:
- Generic files give AI systems no basis to prefer one brand over a competitor in the same category
- Customized files that include hero product descriptions, use-case guidance, and canonical page links give AI systems information they can actually use in a recommendation
- Brands that customize early establish clearer AI-facing signals before the gap between generic and specific content defines the entire competitive gap
Why Customization Is More Urgent Now That the Files Exist
Before Shopify generated these files automatically, brands could reasonably defer. Creating an llms.txt from scratch required effort, the convention was still emerging, and agentic shopping was a future use case rather than an active one.
That calculus has changed on multiple fronts. The files already exist on every domain. AI crawlers can discover them. Shopify officially supports customization via Liquid templates in the theme code editor. And with Shopify's Agentic Storefronts now discoverable in ChatGPT, Microsoft Copilot, AI Mode in Google search, and Gemini, the file is more likely to be fetched than it was six months ago.
The point is not that AI agents are completing purchases for every customer today. The urgency is narrower: a generic file may become the default machine-readable summary of a store.
Brands that customize Shopify’s AI discovery files earlier give AI systems clearer product and brand signals. The window to establish that clarity is now open.
Are These Files for Agents, Crawlers, or AI Search Engines?
Shopify's naming convention points toward agentic commerce, and that is the direction the platform is building toward. But most consumer shopping journeys today are not fully autonomous agent purchases. The more immediate use case is AI retrieval: AI systems fetching, summarizing, comparing, and recommending products in response to natural language queries.
Shopify's own data shows a 8x YoY increase in AI-referred traffic in Q1 2026, with an 13x increase in AI-attributed orders over the same period. Those conversions are coming from AI-assisted searches, not fully autonomous agents. The files matter now for the same reason structured data has always mattered in search: they help non-human systems understand what exists, why it is relevant, and what to recommend.
The practical framing: today, these files help AI systems understand and retrieve accurate brand and product information. Near term, they help shopping assistants navigate catalogs and answer comparison queries. Longer term, they become part of the infrastructure for agentic commerce as defined by Shopify's Universal Commerce Protocol.
The target for customizing Shopify’s AI discovery files are not just fully autonomous shopping agents. Brands should customize them for any AI crawler trying to understand what the store sells and why it is different.
What to Actually Put in Shopify’s AI Discovery Files
The goal is specificity and accuracy. A customized llms.txt should address each of the following:
Brand summary: A clear, factual description of what the brand sells, who it serves, and the specific differentiation that separates it from category competitors. This is not the same as homepage copy. It is a description written for an AI system that needs to categorize and position the brand accurately in comparison queries.
Priority products: Hero products first, with canonical product names, URLs, use cases, key benefits, and differentiating attributes. An AI system asked "what is the best [category] product for [use case]?" needs to know which product to surface. The default Shopify file offers no guidance on this question.
Product selection guidance: This is context for comparison queries. Which product suits which customer profile, what the decision criteria are, and how the product lines differ from each other. This is the content that enables AI systems to recommend products accurately rather than returning a generic category result.
Supporting content: Link to the important pages that back up what the file says about the brand and its products: reviews, FAQs, comparison guides, ingredient or material pages, methodology pages, or any content that substantiates how the brand describes itself. If the file says something about the brand, there should be a page on the site that confirms it. AI systems that detect inconsistencies between the discovery file and the actual pages discount the authority of both.
Canonical pages: Product detail pages, collection pages, comparison guides, FAQs, and policy pages. These are the pages an AI system should treat as the authoritative source for brand information. PDPs are the primary unit of evaluation when AI systems compare products at the SKU level, and the discovery file should route AI systems to them directly.
Freshness signals: Current product launches, discontinued SKUs, major product or description updates, and any changes in availability or pricing. Stale information in a discovery file produces stale recommendations.
What Not to Put in Shopify’s AI Discovery Files
The file should route AI systems to canonical information, not instruct or manipulate them.
Avoid keyword stuffing, information that contradicts what your site actually says, and generic homepage copy. Avoid dumping every product variant without hierarchy or context. Avoid contradictions with what the PDPs, schema markup, and product feeds actually state. AI systems cross-reference what the file asserts against what the pages confirm, and inconsistencies reduce authority for both. Contradictory signals are among the most common reasons e-commerce brands fail to get cited in AI answers.
The file is not a ranking lever. It is a structured brief that gives AI systems the accurate context they need to recommend the brand correctly.
How This Fits into a Broader AI Search Strategy
The discovery files are one layer of a broader AI search readiness framework. A customized llms.txt cannot compensate for thin PDPs, weak schema, or product content that does not hold up under scrutiny. When the rest of the site is strong, it helps AI systems find the right facts faster. The full stack includes product schema, merchant feed data, FAQs, reviews, and external authority signals. Improving AI product discovery requires all of these working together. The discovery files are part of that infrastructure, not a substitute for it.
Conclusion
Shopify has made AI discovery accessible for every merchant on the platform. The files exist, they are discoverable, and customization is officially supported. That also means the baseline has been commoditized. Every brand starts with the same file.
The brands that benefit will be the ones that replace that baseline with accurate, current, brand-specific guidance. The AI-facing surface is live, it is indexed, and the window to establish clear signals ahead of competitors is open now.
If your store is on Shopify, visit your own /agents.md, /llms.txt, and /llms-full.txt. Read what they say. Ask whether that file accurately reflects the products, descriptions, and buying guidance you want AI systems to associate with your brand. If the answer is no, the next step is to write a file that does.
—
Interested in expert authoring of the Shopify llms.txt, llms-full.txt and agent.md files? Stellar provides AEO services to e-commerce brands that include authoring of AI discovery files, in addition to assessment and recommendations across content, structured data, authority signals, indexability and recency. Write to us at contact@stellar-ai.co.