Does llms.txt Help a Shopify Store Get Cited by AI?

Quick answer: Shopify already serves a default llms.txt, and a matching agents.md, on standard stores, so a standard store has one without you doing anything. It is a generic set of instructions for shopping agents, not a map of your best products, and I could not find evidence that it gets a store cited more often. Google says Google Search ignores these files, SE Ranking’s study of nearly 300,000 domains found no effect on citations, and Ahrefs found that 97% of valid files got no requests at all in May 2026. Check what your store serves, edit it only if you will keep it accurate, and put the real effort into crawl access and correct pages (all as at 30 October 2026).

If you run a Shopify store, someone has probably told you that you need an llms.txt file “so ChatGPT can find you”, and that it costs a monthly fee. Then you search it and find guides that disagree: some say Shopify does not let you add one, some say Shopify creates one for you, some say it is a theme template, some say upload it to the root.

They disagree mostly because the platform changed during 2026 and the guides were written at different points. So I did not rely on them. On 30 October 2026 I opened the files on real Shopify stores, then read Shopify’s documentation, Google’s guidance, the llms.txt proposal and the two large studies that tried to measure whether the file does anything.

This post is the practical result: what Shopify serves by default, what is documented, what is not shown, and a short check for your own store. Our broader post on GEO myths Shopify founders still believe grades llms.txt as one of seven claims; this one goes deeper on the practical side.

What llms.txt is meant to be

llms.txt is a proposal, not a formal standard. The proposal page says it was written by Jeremy Howard, first published on 3 September 2024 and last modified on 10 August 2026. It describes a Markdown file at /llms.txt that gives AI agents short background on a site and links to more detailed Markdown pages. It says the file is mainly expected to be useful for inference (an agent answering something right now), not for training models.

The format it describes is specific:

• An H1 with the site or project name. This is the only required part.
• A short summary in a blockquote.
• Optional sections of detail.
• H2 sections containing lists of links to the pages that matter, each with a short note.

The proposal page says thousands of sites publish one, that documentation platforms can generate one automatically, and that OpenAI, Anthropic and Google’s Gemini publish one for their developer documentation. It says the file is designed to coexist with robots.txt and sitemaps, which do different jobs.

Notice what the page does and does not say. It says the same structure can work for a business outlining its structure and policies, and that chat assistants with search read pages. It does not say that a shopping assistant reads a store’s file to decide which brand to name. A lab publishing a file for its own developer docs is not the same thing as an answer engine using your file to pick products.

What Shopify serves on your store by default

Shopify’s developer changelog entry of 28 May 2026 says your store includes a default agents.md file at /agents.md, and that /llms.txt and /llms-full.txt point to the same content by default. Shopify’s template documentation says it manages an agents.md for every store, describes /agents.md as the canonical description of a store for AI agents and shopping assistants, and describes /llms.txt as an alternate URL that mirrors it.

So on a standard Shopify store you do not need to create an llms.txt file. It already exists.

Test it yourself in two minutes

Open these three addresses on your own domain:

• yourstore.com/llms.txt
• yourstore.com/agents.md
• yourstore.com/llms-full.txt

I did exactly this on 30 October 2026, and here is what came back.

In total I checked nine standard storefronts (including an OS 2.0 demo store, Kylie Cosmetics, Death Wish Coffee and Tanorganic) and all nine served the default on all three addresses. The two Hydrogen stores are different. Hydrogen is Shopify’s headless toolkit, and both stores’ response headers name it. One served its own file and no agents.md, and the other served nothing. I did not find Shopify documentation on what headless storefronts serve, so if yours is headless, test the URL and ask your developer, because the default may not apply to you.

One small thing to check: Shopify’s documentation says the file is served at your bare primary domain, with no locale or Markets prefix, and has no localised version. If you sell into the US from an Australian store using Shopify Markets, there is one file, not one per market.

What is actually in the default file

I compared the file across the nine stores. After you swap the store name and web address, the text is the same on all of them apart from which policy links apply (two stores had no shipping policy link) and the capitalisation of three policy labels on one store. It runs to about 760 words, and the /llms.txt and /agents.md versions differ by one sentence, the one saying which URL is canonical.

It is not a description of your brand. Its sections are:

1. A recommendation to use the Shop skill, which the file describes as offering personalised discovery and accelerated checkout, and as working across every Shopify store.
2. How an agent can connect to the store using the Universal Commerce Protocol (UCP), which the file describes as covering product discovery, carts and checkout, including the store’s MCP endpoint.
3. How a browser-based agent should use the store’s WebMCP tools.
4. Interaction rules: buyer authorisation, agent identification and rate limits.
5. Read-only browsing: public product, collection and sitemap addresses, in generic form.
6. Links to your privacy, terms, refund and shipping policies.
7. A “Platform” section that describes Shopify and links to start a store.

There is no list of your best sellers, no summary of what you sell and no description of who your customers are. Shopify’s documentation says the generated file deliberately leaves out contact details.

Read as a whole, the file is about agents that want to transact with your store, not about answer engines deciding who to cite. That is a different job from the one the llms.txt proposal describes, and the file borrows the name. Parts of it also promote the Shop skill and Shopify itself, which is worth knowing when someone tells you your “AI file is already optimised”.

On the same demo store, the comments at the top of the robots.txt point agents to /agents.md and the UCP discovery address. The robots.txt had no rules naming AI crawlers such as GPTBot or OAI-SearchBot, which matters in the check below.

Why the guides disagree

The dates explain most of it.

• July 2025. A thread on the Shopify developer community began with a merchant saying Shopify did not allow a file in the root directory. Over the following months, people suggested workarounds: upload the file under Content > Files and redirect to it, put a Cloudflare Worker in front of the store, or use an app proxy. Others doubted the redirect would work.
• 8 May 2026. A community poster in that thread wrote that it was now solved and worked like robots.txt, with a templates/llms.txt.liquid file.
• 28 May 2026. Shopify’s changelog documented the default agents.md, the mirroring to /llms.txt and /llms-full.txt, and three optional templates. The 8 May post is earlier than the changelog, so templates may have been available a few weeks before the announcement; I did not find an earlier Shopify source.

A guide written in 2025 may describe a setup that no longer applies. Another poster in the same thread, in September 2026, describes the native default as a generic mirror of agents.md. That poster develops an app that does the job, so weigh the comment accordingly, but my own comparison above agrees with the description.

Can you change it?

Yes, through a theme template, and Shopify tells you not to bother unless you have a reason. The documented rules are:

• You add a template under Online Store > Themes > Edit code. The three names are agents.md.liquid, llms.txt.liquid and llms-full.txt.liquid. They are optional and not part of a theme by default.
• llms.txt.liquid controls only /llms.txt. agents.md.liquid controls /agents.md and is the fallback for the other two. If no template exists, the Shopify-generated default is used.
• The template cannot be a JSON template, and it has to be named exactly.
• The Liquid context is restricted. Only the request and agents objects are available. Objects such as shop, collections and articles are not injected, so you cannot automatically list your products. The agents object gives you the store name, store URL, currency, UCP and MCP addresses, supported UCP versions and the sitemap address.
• Once you add a template, Shopify says you take responsibility for keeping the content current.
• Shopify warns against outputting private details, such as contact emails or phone numbers, because the file is cached broadly and served to every agent.

Two practical notes from my reading of those rules (my inference, not Shopify’s advice). First, Shopify recommends agents.md.liquid for one change that covers all three addresses. But if all you want is a short, human-written summary at /llms.txt, a separate llms.txt.liquid leaves Shopify’s agent-shopping instructions at /agents.md untouched. Second, because the file has no access to your catalogue, anything you write about products is hand-maintained, and a stale file is a worse outcome than the generic default.

What is proven, and what is not

Here is the evidence as I read it, with the primary source for each line.

Google

Google’s guide to optimising for generative AI features on Google Search (last updated 10 July 2026) names llms.txt in a section on things you can ignore. It says you do not need new machine-readable files, AI text files, markup or Markdown to appear in Google Search, including its generative AI features, because Google Search does not use them. It adds that it is fine to create them for other services, and that doing so will neither harm nor help your visibility or rankings in Google Search, as Google Search “ignores them”.

Google’s separate AI features page (last updated 10 December 2025) does not name llms.txt but says the same thing more generally: there are no additional requirements for AI Overviews or AI Mode, a page needs to be indexed and eligible to show with a snippet, and you do not need new machine-readable files, AI text files or special schema.

Chrome treats the file differently, and it is fair to say so. Chrome for Developers documents an llms.txt check in Lighthouse’s agentic browsing audits (page last updated 5 May 2026). It calls the file an emerging convention, says that without it agents may spend more time crawling a site to understand its structure, and says providing it is optional at the moment: a missing file is marked not applicable. That is documentation for a testing tool, not evidence of citations.

Google’s John Mueller said much the same as the Search documentation in two Bluesky posts I read directly. On 17 June 2025 he replied to a post about turning on llms.txt, writing that no AI system currently uses it. On 20 January 2026 someone asked whether Google hosting an llms.txt on its Gemini API docs was an endorsement, and he answered no. Those are one Google employee’s personal comments, not documentation, and the first is over a year old. Ahrefs’s data below shows some AI bots do fetch the file, so read his 2025 line as a view on use, not a statement about fetching.

Other assistants: what I could and could not find

For ChatGPT, the page I read is OpenAI’s crawler documentation. The controls it describes are robots.txt rules for its crawlers: allow OAI-SearchBot if you want to appear in ChatGPT search results, and disallow GPTBot if you do not want your content used to train its models. It says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, and that robots.txt changes can take about 24 hours to take effect for search. The only mention of llms.txt on that page is a link to OpenAI’s own documentation index. The page does not describe llms.txt as a way to be included.

That is one page from one company. I did not read crawler or search documentation from Perplexity, Anthropic or Microsoft for this post, so I make no claim about them either way. If an assistant’s maker says its tool reads a store’s llms.txt to choose sources, that will be in its own documentation.

The two big studies

SE Ranking (page dated 15 October 2026) looked at nearly 300,000 domains. Of those, 10.13% had an llms.txt file. It looked at domain-level citation frequency across sampled AI responses, using Spearman correlation, an XGBoost regression model and SHAP analysis. It reports no correlation between the file and citations, and says removing the llms.txt variable actually made its model more accurate. Its conclusion is that the file does not affect how AI systems see or cite content today. It also says the results are specific to its model and dataset and can change, calls adding the file a low-effort way to prepare for the next wave of AI indexing (it adds that the file “might be essential” later, which is a prediction, not a finding), and says to treat it as an optional experiment while fixing the technical barriers a file cannot fix. Its methodology section does not say which AI tools were sampled.

Ahrefs (published 15 June 2026) took a different approach: server-side traffic. It looked at 137,210 domains in its web analytics that had traffic in May 2026. About 28% (38,360) had a valid llms.txt, meaning the address returned a Markdown file and an HTTP 200. Of those, 97% received no requests for the file at all in that month. Across the files that did get requests, there were only about 22,000 requests in total. Ahrefs sorted the requesters:

• Bots made 96% of requests.
• SEO audit tools made 21.7%, including Ahrefs’ own crawlers at 10.6%.
• AI bots combined made 19.5%: agents and agentic tools 10.5%, training crawlers 5.3%, assistants 2.5% and retrieval bots 1.1%.

Ahrefs concluded that, for AI search visibility, the cons outweigh the pros right now. It says the closest thing to an intended audience in its data is coding agents.

What these studies cannot tell you

They do not prove the file does nothing.

• Both are observational. SE Ranking looks for a pattern between having the file and being cited. A missing pattern is not proof of no effect, especially for small effects or for specific tools.
• They measure different things. One looks at citations, the other at fetches. A fetch is not a read, and Ahrefs says its figures are a ceiling on actual use.
• Each has stated limits. Ahrefs says its customers skew towards technical and SEO-aware sites, so the 28% adoption figure is an upper bound, and it did not check files against the specification.
• Timing and platform. Ahrefs counted requests across May 2026, and Shopify documented its default on 28 May, so the default may have been live for little or none of that window. Neither study breaks results out by platform, so neither says anything specific about Shopify’s default. SE Ranking does not say when its data was collected. Both were published by SEO software companies, so treat them as vendor research, not independent research.
• Agents are not answer engines. An agent working through a store on a buyer’s behalf may behave differently from a model composing an answer. I found no data either way for Shopify’s agents.md.

So the honest position is this. For Google Search, there is documentation saying it does not matter. For other assistants, the best public measurements show little use and no clear citation benefit, and nothing shows a gain. Nobody has shown a harm either, provided the file is accurate. Do not claim citation gains from it, and be wary of anyone who does.

A 20-minute check for your store

None of this needs an app.

1. Open the three URLs (/llms.txt, /agents.md, /llms-full.txt) on your primary domain. Confirm they load and show your store name and address, not a staging address or an old domain.
2. Look for custom templates. In Online Store > Themes > Edit code > Templates, check whether agents.md.liquid, llms.txt.liquid or llms-full.txt.liquid exist. If none do, you are on the default. If one does, find out who added it and when.
3. Check for older workarounds. If you or a developer set up a redirect from /llms.txt (the 2025 method) or installed an app that makes the file, open the URL and see what actually loads. I did not test which wins when several exist, so verify it rather than assume.
4. Read the file as a customer would. If you did write custom content, check that every statement matches your live policies, shipping countries and prices. An assistant may repeat what the file says. A stale or exaggerated line is your problem to fix.
5. Check your robots.txt. The control that is documented for ChatGPT search is robots.txt. Open yourstore.com/robots.txt and look for rules naming OAI-SearchBot, GPTBot, Googlebot or other crawlers. On the demo store I checked there were none. If your store has them, make sure they are intentional. Our guide to Shopify’s robots.txt explains what you can and cannot edit.
6. Check your apps. If you pay monthly for an llms.txt app or an “AI visibility” line item, ask what exactly it delivers. If the answer is “more citations”, the evidence above does not support that promise.
7. Put a re-check in your calendar. Shopify introduced this in 2026 and the documentation may change. Check the three URLs again in a quarter.

Should you customise it?

If you do write one, an hour is the right size of project. Use the proposal’s shape: your name, a short summary and short lists of links. Here is an illustrative skeleton for llms.txt.liquid, using the objects Shopify documents. Replace every bracketed part and try it on a duplicate theme first.

# {{ agents.store_name }}

> [One or two sentences: what you sell, who it is for, and which countries you ship to.]

## Main collections
- [Collection name]({{ agents.store_url }}/collections/[handle]): [one line on what is in it]

## Policies
- [Shipping and delivery]({{ agents.store_url }}/policies/shipping-policy)
- [Refunds and returns]({{ agents.store_url }}/policies/refund-policy)

## Sitemap
- [Sitemap]({{ agents.sitemap_url }})

Say only what you would put on a product page and could back up. Leave out superlatives such as “Australia’s best” and private contact details, as Shopify advises. It is a summary for a reader, not a place to repeat keywords.

For reference, prism3.agency publishes an llms.txt file on its own site, which is not a Shopify store. We make no claim that it earns citations.

What to spend the time on instead

Google’s own guidance on AI features comes down to the same boring list that applies to ordinary search. These are the things that are documented, and where I would put the hour you saved:

• Pages that are indexed and eligible for a snippet. Google says that is the technical requirement for AI Overviews and AI Mode. If important pages are missing from Google, fix that first. See our guide to pages Google crawls but does not index.
• Important content as text. Google’s AI features page lists making sure important content is available in textual form. If prices, sizing or delivery times appear only inside images, add them as text.
• Structured data that matches the visible page. Google says it should match the visible text and that there is no special schema to add. Our post on schema markup and AI citations covers what to check.
• Crawling that is actually allowed. Google’s page lists allowing crawling in robots.txt and by any CDN or hosting infrastructure. See step 5 above.
• Accurate product and business details. Google’s generative AI guide has a section on ecommerce details and names Merchant Center feeds, and the AI features page says to keep Merchant Center and Business Profile information up to date.
• Content that answers buyer questions. Our guide to ranking in ChatGPT covers the parts of this that are about the content itself.

If your interest is agents that shop for people rather than answer engines that cite them, that is a separate topic. We cover it in Shopify agentic storefronts and, for the Australian view, AI shopping in Australia.

How would you know if it worked?

You mostly cannot isolate it. If you add a file and then see more visits from AI tools, many other things changed in the same weeks. What you can measure is whether visitors arrive from AI tools and whether they buy. Set that up first with tracking ChatGPT traffic in GA4. An agency or vendor claiming credit for the file should be able to name which system fetched it and when.

Key takeaways

• On standard Shopify stores, /llms.txt, /llms-full.txt and /agents.md already exist. I confirmed it live on nine stores on 30 October 2026.
• The default is generic agent-shopping instructions, the same on every store apart from name, address and policy links. It is not a summary of your brand or your products.
• Google says Google Search ignores llms.txt files, so having one will neither help nor harm your Google Search visibility.
• SE Ranking found no effect on citations across nearly 300,000 domains, and Ahrefs found 97% of valid files got no requests in May 2026. Neither proves the file does nothing, and neither tested Shopify’s default.
• If you edit the file, use a llms.txt.liquid template, keep it to one page, put nothing private in it, and own the upkeep. Do not pay monthly for citations.
• Spend the time on crawl access, indexed pages, readable text and accurate product data, which are documented.

FAQs

Does Shopify create an llms.txt file automatically?

On standard storefronts, yes. Shopify’s changelog of 28 May 2026 says your store includes a default agents.md and that /llms.txt and /llms-full.txt point to the same content. I confirmed it on nine stores on 30 October 2026. Headless stores may differ.

How do I edit my Shopify llms.txt?

Go to Online Store > Themes > Edit code and add a template named llms.txt.liquid in the Templates folder. Use agents.md.liquid instead if you want one change to cover all three addresses. Shopify says you are responsible for keeping any template current.

Does llms.txt help me appear in Google AI Overviews?

No. Google’s guide says Google Search itself does not use these files, and that creating one will neither harm nor help your visibility or rankings in Google Search. A page appears in AI Overviews and AI Mode by being indexed and eligible for a snippet, according to Google’s AI features documentation.

Will ChatGPT or Perplexity read my llms.txt?

I found no primary source showing that ChatGPT uses a store’s file to choose what to cite, and I did not check Perplexity’s documentation. OpenAI’s crawler documentation describes robots.txt controls for OAI-SearchBot and GPTBot, not llms.txt. Ahrefs’s server data shows some AI-bot fetching, at 19.5% of the roughly 22,000 requests that files received, and SE Ranking found no effect on citations. That does not prove no tool ever reads it, but it gives you no reason to expect a gain.

Should I pay for a Shopify llms.txt app?

Not for citations, because nothing I read shows that outcome. Shopify serves a default file and lets you edit it with a template, so you do not need an app to have one. An app may still save you time if you want a maintained, richer file and the price is low, but ask what exactly it does, and check which file your URL serves after you install it.

Does llms.txt replace robots.txt or my sitemap?

No. The proposal says robots.txt and llms.txt have different purposes and that a sitemap is not a substitute for llms.txt, so they sit side by side. SE Ranking notes that adding llms.txt does not override restrictions in robots.txt. If you want AI search crawlers such as OAI-SearchBot to reach your pages, OpenAI documents robots.txt as the place to allow them.

Want a second opinion on whether AI search can read your store?

Our GEO service is delivered alongside SEO, and the basics in this post (can your pages be crawled, indexed and read, and is your product data accurate) sit inside that work. We make no promise that any AI assistant will cite you. If you would like a second pair of eyes first, book a free 30-minute diagnostic or tell us about your store. We look at your store, and at what the AI assistants say about your category, before we recommend anything. No pitch.