What AI Assistants Actually See When They Read Your Shopify Store

You can argue all day about AI shopping in the abstract, or you can go look at exactly what an agent reads when it encounters your store. The second is more useful, and it takes about ten minutes. This is the walkthrough: the precise files an AI assistant reads from your Shopify store, how to view them yourself, and how to fix what gives a weak impression.

Step one: read your own agents.md

Shopify now auto-generates discovery files for every store. Visit yourstore.com/agents.md and read it. If you try /llms.txt, it likely redirects you to /agents.md, which is Shopify standardizing AI discovery around its commerce protocol.

The two files do different jobs. The llms.txt half is a curated reading list, pointing an AI system to your best product, FAQ, guide, and policy pages. The agents.md half is the operating manual, telling an agent how to search, cart, check out, and which commerce protocols you support. Read together, they're the agent's first impression of your store. We break down each in agents.md for Shopify and llms.txt for Shopify.

What you're looking for: does the brand summary actually describe what you sell, or is it generic boilerplate? The default is serviceable but anonymous. Injecting a real brand summary at the top, so the agent learns who you are before it reads the checkout rules, is one of the highest-value edits available, and a default left untouched can quietly work against you, which we cover in the default agents.md that's killing your brand.

Step two: read a product the way an agent does

Open one of your product pages and view its structured data (right-click, view source, search for application/ld+json, or use a schema validator). This is where the agent gets the facts: the price, the rating, the availability, the product type.

Now ask yourself the agent's question. If a shopper said "find me something like this," would this data let the assistant describe your product accurately and confidently? Agents pull titles and descriptions close to verbatim. An over-branded title ("The Summit Pro") and a description full of adjectives give the agent nothing concrete to repeat, so it either skips you or describes you vaguely. That's the root cause behind most "the AI got my store wrong" complaints. The fix is concrete, quotable product data, covered in descriptions that AI can parse.

Step three: check the door is unlocked

None of this matters if the agent can't fetch the page. Open yourstore.com/robots.txt and confirm the AI crawlers (OAI-SearchBot, PerplexityBot, and others) aren't blocked. It's a thirty-second check that silently sinks stores that skip it. We walk through it in AI crawlers and your Shopify robots.txt.

What the ten-minute audit usually reveals

Run those three steps and a pattern almost always emerges. The crawler access is usually fine. The discovery files exist but read generically. And the product data is the weak link: titles too branded to match queries, descriptions written for human eyes, and structured data that's incomplete or missing on the pages that matter. That's not a failure of effort. It's that the entire catalog was built for human readers, and the agent reader showed up later.

Closing the gap, and keeping it closed

The audit tells you what to fix. Two things keep it fixed. First, the structured data and discovery files have to stay accurate as your catalog moves, which is the standing job AgentReady Signal handles: it publishes and maintains the schema and AI-discovery files so what an agent reads stays current. Second, the content an agent reaches for, your guides, comparisons, and answers, has to keep being produced in a form a model can quote, which is the job of AgentReady Reach: it drafts posts grounded in your real catalog and brand voice, publishes them to your Shopify blog, and measures the lift in your own Search Console. We get into why content for a model reads differently in writing for a reader that's a model.

Start with the look, though. The free Shopify AI readiness checker runs this same audit on any storefront in about a minute and tells you what an assistant sees and what to fix first. You can't improve the impression your store makes on an agent until you've read it the way the agent does.