How to audit a Shopify store for AI search and shopping agents

Key takeaways

What to remember

• AI-search readiness is broader than adding a single llms.txt file.
• Preserve Shopify-generated agent discovery links before overriding llms.txt.
• Product, collection, policy, and FAQ pages need to answer buyer questions in rendered HTML.
• Agents need clear routes to search, browse, contact, shipping, returns, and checkout rules.
• Audit the live storefront response, not only theme files or admin settings.

AI shopping is still early, but the shape is clear enough to audit against.

A shopper may ask ChatGPT, Perplexity, Gemini, or another assistant for a product recommendation. The assistant may crawl the web, read structured data, inspect store policies, compare collections, and send the shopper to a product or checkout page. On Shopify, that path is starting to include files like /llms.txt, /agents.md, and agent discovery sitemaps.

That does not mean every merchant needs to panic-build an agentic commerce stack this week. It does mean the store should be legible to software that is not browsing like a person.

A good Shopify AI-search audit answers one question: if an AI system lands on this store, can it understand what is sold, who it is for, how buying works, and which pages deserve trust?

The short version

For Shopify merchants, AI-search readiness is not one file. It is the combined state of:

• llms.txt and agent-discovery files
• product and collection pages
• structured data
• sitemap coverage
• policy and FAQ pages
• internal links
• checkout and support instructions
• the actual rendered HTML that crawlers can see

The last point matters. Theme settings, app settings, and Liquid templates are not enough. Audit the live response.

Start with the live storefront

Do not begin in the Shopify admin. Begin like a crawler.

Open the public store and check the pages an assistant is most likely to use:

• homepage
• top collections
• best-selling or representative products
• search results
• contact page
• shipping policy
• refund policy
• privacy policy
• FAQ or help center
• blog or buying guides, if the store has them

The basic test is boring, which is why it gets missed: can you tell what the store sells without relying on images, JavaScript interactions, or brand vibes?

A human can infer a lot from photography. An agent needs text, links, schema, and predictable page structure.

Check llms.txt without treating it like magic

/llms.txt is a proposed convention for giving large language models a short map of a site. Shopify has been testing auto-generated versions on some storefronts, and merchants can override the response with a theme template such as templates/llms.txt.liquid.

That override is useful, but it is also easy to get wrong.

If Shopify is already generating a default file, it may include links to /agents.md, /.well-known/ucp, /api/ucp/mcp, the sitemap, browse pages, and search patterns. Replacing that with a thin custom note can remove useful agent-discovery hints.

A safer audit asks:

• Does /llms.txt exist?
• Does it describe the store plainly?
• Does it link to important collections, products, guides, and policies?
• Does it preserve Shopify’s generated agent/developer links where present?
• Does it avoid stuffing every product URL into one file?
• Does it point agents to canonical sources instead of duplicate or filtered URLs?

A strong llms.txt should be a map, not a sales pitch.

Check agents.md and discovery routes

/agents.md is more operational. Where it exists, it can explain how an agent should browse products, use search, understand checkout rules, and handle payment approval.

Most merchants will not hand-author this file yet. Shopify may own the route on stores where its agentic-commerce tests are active. That is fine. The audit job is to record what exists, what it says, and whether the rest of the store supports it.

Check:

• /agents.md
• /sitemap_agentic_discovery.xml
• /llms-full.txt, if exposed
• /.well-known/ucp, if exposed
• /api/ucp/mcp, if exposed

Do not overstate what these files prove. A route existing does not mean AI orders are being attributed cleanly, and it does not mean the merchant can see agent metadata in normal Shopify order webhooks. Treat these as discovery signals.

Audit the product pages like an answer engine would

Product pages are where many Shopify stores look fine to humans and weak to machines.

A good AI-shopping product page should make the core buying facts explicit:

• product name
• price and availability
• variants
• size, dimensions, materials, ingredients, compatibility, or fit
• shipping constraints
• returns or warranty notes
• trust signals
• reviews, if present and legitimate
• structured product data
• links to related guides or policies

Do not hide essential information inside images, accordions that render late, or app widgets that disappear for crawlers.

The page should also answer the questions a buyer would ask before purchase. For apparel, that might be sizing and returns. For supplements, ingredients and usage. For electronics, compatibility and warranty. For home goods, dimensions, materials, care, and delivery.

If those answers exist only in customer support macros, they are not helping AI search.

Audit collection pages as category explainers

Collection pages often get treated as product grids with a two-sentence intro. That is thin for search and even thinner for agents.

For the most important collections, check whether the page explains:

• what belongs in the category
• who the category is for
• how to choose between products
• what filters or variants matter
• which buying guide or FAQ page answers deeper questions

This does not mean turning every collection into a 2,000-word article. It means adding enough useful text and internal links that an agent can understand why the collection exists.

If a store sells running shoes, “Men’s Running Shoes” should not only be a grid. It should tell a buyer how to choose by terrain, support, distance, fit, and return policy.

Check schema, but do not stop at schema

Product schema helps crawlers parse price, availability, ratings, and product identity. It should be present and valid.

But schema is not a substitute for page content.

Audit both:

• JSON-LD or microdata for Product, Offer, AggregateRating, BreadcrumbList, and Organization where relevant
• visible page text that supports the same facts

If schema says one thing and the page says another, that is not readiness. That is ambiguity.

Also check whether multiple apps are injecting duplicate or conflicting schema. Shopify stores often accumulate overlapping SEO, review, and product apps. An AI system may not know which version to trust.

Make policies easy to find and easy to parse

Shopping agents need buying rules. Shipping, refunds, privacy, contact, subscriptions, warranties, delivery restrictions, and payment constraints all matter.

A policy page should be crawlable, linked, and specific enough to answer common buyer questions.

Weak policy pages create avoidable friction:

• “Contact us for details” instead of clear shipping timelines
• return windows buried in legal copy
• subscription terms hidden behind app widgets
• no page for warranty or fit guidance
• policy links missing from product pages

For AI search, policy clarity is part of conversion. If an assistant cannot confidently answer “Can I return this?” or “Does this ship to my state?”, it may recommend a different store.

Look for rendered HTML gaps

A Shopify theme can look complete in the browser and still be thin in the source that crawlers receive.

Check the rendered HTML for:

• title and meta description
• canonical URL
• crawlable product details
• crawlable collection copy
• internal links
• FAQ content
• schema scripts
• policy links
• status codes and redirects

This is where many audits become too theoretical. The store either serves the content or it does not.

Prioritize fixes in the right order

Do not start with the fanciest agentic-commerce acronym.

Fix the basics first:

1. Make the main pages crawlable and specific.
2. Publish or preserve a useful /llms.txt.
3. Keep Shopify’s generated agent discovery links where present.
4. Improve product and collection copy where buying questions are missing.
5. Validate product schema and remove conflicts.
6. Link policies from the places buyers and agents need them.
7. Add FAQs only where they answer real pre-purchase questions.
8. Re-test the live responses.

This order keeps the work practical. A merchant gets value even if AI shopping evolves slowly, because the same fixes help search engines, buyers, support teams, and conversion rate.

Where Honeybound fits

Honeybound audits Shopify stores for this exact layer: not just whether a file exists, but whether the store gives AI systems a reliable path through the catalog, policies, and purchase context.

The audit looks for things like LLM file coverage, agent discovery routes, product and policy clarity, schema quality, internal link paths, and the risk of overriding Shopify’s defaults incorrectly.

If the report finds missing or weak coverage, the fix is usually concrete: publish a safer llms.txt, strengthen collection pages, expose policy answers, clean up schema, or add the internal links that help agents reach the right source.

Run the audit here: Shopify AI commerce readiness audit.

Final check

AI shopping will not reward stores for having the longest file or the most buzzwords. It will reward stores that are easy to understand.

That is the useful standard for an audit. Can a machine read the same facts a careful buyer needs before ordering?

If the answer is yes, the store is in a better position for AI search, agentic commerce, and ordinary human shoppers too.

Frequently asked questions

What is a Shopify AI-search audit?

It is a review of whether AI crawlers and shopping agents can understand a Shopify store's catalog, policies, structured data, discovery files, and checkout path.

Is llms.txt enough for Shopify AI readiness?

No. llms.txt helps point AI systems to useful pages, but the pages themselves still need clear content, schema, policy information, and internal links.

Should Shopify merchants override the default llms.txt file?

Only with care. If Shopify is generating useful agent links, preserve those links before adding merchant-specific guidance.

What should merchants fix first?

Fix missing discovery files, thin collection pages, weak product schema, hidden policy pages, and confusing checkout or support instructions before chasing more advanced agentic-commerce work.