Does your Shopify store need an llms.txt file, or is good product data enough?
Your store may have an `llms.txt` file already.
If you heard that AI search requires one, it is worth checking before installing anything.
Shopify automatically serves three discovery files for Shopify stores:
- `/agents.md`
- `/llms.txt`
- `/llms-full.txt`
Shopify calls `/agents.md` the canonical URL. The two `llms` paths are there for older crawlers that expect them.
That means most Shopify merchants do not need an app just to create an `llms.txt` file.
The more useful question is simpler: can an AI system find, read, and trust the facts about your products?
Check your store in under a minute
Open this URL in a browser, replacing `yourstore.com` with your domain:
`https://yourstore.com/agents.md`
Then try:
`https://yourstore.com/llms.txt`
You should see a Markdown document rather than a normal store page.
If it loads, Shopify is already handling the basic discovery file.
If you use a password-protected development store, test this after your storefront is public. A file that exists behind a password is not much help to a crawler outside your store.
What the file does, and what it does not do
An `llms.txt` file is a proposed format for helping language models find important pages on a site. Its basic shape is simple: a site name, a short overview, and links to useful pages.
The format itself does not make product details accurate. It does not update an out-of-stock variant. It does not add dimensions to a product missing them. It does not tell a shopping system which option costs more.
That distinction matters because Shopify treats these discovery files separately from Shopify Catalog.
Catalog is Shopify's authoritative product-data feed for its agentic channels. It can include product titles, descriptions, options, images, prices, availability, and other attributes. Shopify also says Catalog keeps pricing and inventory updated.
Think of the discovery file as a sign at the entrance.
Your product data is what is on the shelf.
A good sign cannot fix the wrong price tag.
Good product data is the first job
Say your store sells insulated water bottles.
A title such as `Trail bottle` gives a system very little to work with. It does not say capacity, material, lid type, insulation time, or whether it fits a cup holder.
A better product record might include:
- A clear product title
- Capacity, such as 750 ml
- Material, such as stainless steel
- Colour and size options
- Lid type
- Care instructions
- What comes in the box
- Current price
- Current availability
- Clear images for each variant
Those details help people too. They also give AI systems fewer gaps to guess through.
Open one of your best-selling product pages now. Ask four plain questions:
- Does the title say what the product is?
- Are its key specs visible without opening five tabs?
- Can a shopper tell which variant is available?
- Does the description answer the question that causes returns?
If one answer is no, start there before worrying about a custom discovery file.
Catalog and sales channels need attention too
Shopify says eligible products can be automatically discoverable through Catalog without separate manual feeds or bespoke integrations.
But eligibility still depends on the product data and sales channels you have set up.
There are also channel-specific paths. Shopify says Google AI Mode and Gemini use the Google & YouTube channel. Meta surfaces use the Facebook and Instagram channel.
That is why a single file cannot be your whole AI-search plan.
Check a few products in Shopify admin:
- Confirm the product is active, not draft.
- Confirm price and inventory are current.
- Check every option and variant.
- Look for empty fields that matter to the product, such as size or material.
- Review the sales channels where the product is available.
Do this first for products that drive most of your revenue. You do not need to perfect an entire catalogue in one afternoon.
Do not confuse robots.txt with Catalog distribution
Some merchants block AI crawlers in `robots.txt`. That choice may make sense for their store.
But Shopify says those rules apply to open-web crawling. They do not stop Shopify from sending product data to agentic storefronts that you have activated through Catalog.
So there are two separate decisions:
- Whether open-web AI crawlers can crawl your storefront
- Whether Shopify can distribute your product data through activated agentic channels
Check your robots rules at:
`https://yourstore.com/robots.txt`
Do not change it just because someone says every store should allow every bot. First decide what you want open-web crawlers to access. Then review your activated Shopify channels separately.
What about Google search?
Google's shopping guidance puts the focus elsewhere: product structured data and Merchant Center feeds.
Google recommends using `Product` structured data, Merchant Center data feeds, or both. It says using both can help maximize eligibility and help verify product information.
For stores aiming at all shopping-result types, Google recommends putting `Product` structured data in the initial HTML. It warns that markup generated only with JavaScript can make crawling less frequent and less reliable when price and availability change quickly.
You can inspect this yourself.
Open a product page in Chrome. Right-click and choose View page source. Search for:
`application/ld+json`
Then look for product facts such as:
- `name`
- `offers`
- `price`
- `availability`
Seeing JSON-LD does not prove every field is correct. It gives you a useful place to start.
We learned a version of this lesson on our own site. We once searched the homepage HTML for our analytics snippet, found nothing, and nearly decided analytics was broken. It was running. The key was in the compiled JavaScript bundle, not the initial page source.
Structured data has the opposite concern for many stores: if important product markup appears only after JavaScript runs, a crawler may not get a dependable view of changing details. Check the source, not only what your browser renders.
When would you customize Shopify's files?
Customization is optional.
Shopify allows merchants to add theme templates for `agents.md`, `llms.txt`, and `llms-full.txt`. If a specific template is not present, Shopify falls back to `agents.md` or to its generated default.
A custom file may be useful when you have a clear reason to control its wording or links. For example, you may want it to point clearly to a buying guide, returns policy, or collection structure that helps explain a complex catalogue.
It is not useful as a substitute for basic product maintenance.
We know how easy it is to overvalue the file itself. When we built our own `llms.txt` builder, we had to handle Shopify catalogue pagination because `/products.json` returns at most 250 products per page. A catalogue read can look complete while quietly missing products.
The file is only as useful as the source data behind it.
That is also why Descriva is not about selling you a text file that Shopify already provides. It audits how ChatGPT, Perplexity, and Gemini can see your store, then helps you work through the product facts and page structure that affect that view. That can include entity-first copy, FAQs, specification lists, JSON-LD schema, Google SEO checks, Search Console work, and indexing work.
The file may be part of the picture. It is rarely the main problem.
Start with the boring, valuable work
For most Shopify stores, the order of work is:
1. Check that Shopify's generated discovery files load. 2. Make key product data complete and current in Shopify. 3. Confirm the right sales channels are set up. 4. Review product structured data and Merchant Center data for Google. 5. Customize discovery files only when you have a specific gap to solve.
If you want a clearer view of what AI systems can actually read on your store, run a Descriva audit and begin with the issues it finds on your highest-priority products.