Shopify shipped a small theme change yesterday that quietly
hands premium merchants the steering wheel of their agentic
commerce surface. As of May 28, 2026, the auto-generated
/agents.md, /llms.txt and
/llms-full.txt on every Shopify store can now be
replaced with theme-level Liquid templates. For studios building
for AI-native shopping, this is the first time the merchant, not
the platform, decides how the brand speaks to an LLM shopping
agent.
What changed
Until yesterday, the agentic discovery files on a Shopify
store were a take-it-or-leave-it default. Shopify began
auto-generating /llms.txt,
/llms-full.txt and /agents.md on every
store in late April or early May 2026, alongside
/.well-known/ucp, /api/ucp/mcp and a
dedicated /sitemap_agentic_discovery.xml. The content
was assembled from store profile data: name, description, contact
details, the public catalogue. Merchants could nudge it by editing
the profile fields, but they could not rewrite it.
The new
changelog
entry changes that. Developers can now drop three Liquid
templates under Online Store > Themes > Edit
code.
The templates/agents.md.liquid template controls
/agents.md and acts as the fallback for the other two
paths. The templates/llms.txt.liquid template
controls /llms.txt only. The
templates/llms-full.txt.liquid template controls
/llms-full.txt only. If a path has no custom
template, it falls back to your agents.md template.
If that template is also missing, the path falls back to
Shopify's generated default.
Why it matters for premium studios in Asia
For premium brands, this is the first lever they have on how
an AI shopping agent talks about them. The default
agents.md tells the agent how to discover
capabilities, search products, create a cart, initiate checkout,
fulfil the order, and complete the transaction. It also enforces
that checkout requires human approval. That covers the mechanics.
It does not cover the brand. It does not say which collections
are the front door, which gift bundles a concierge agent should
suggest first, what the return window looks like for the
Indonesia store versus the Singapore store, or how to introduce a
new product line to a buyer who arrived through a third-party
agent.
For a Jakarta or Manila based studio building a premium DTC storefront, this is a content surface that has been missing from the brief. Until now, the agentic conversation about the brand was happening invisibly, with the platform reading the catalogue and inventing the narration. Shopify just made that narration something a designer and a copy lead can write together. The work looks less like SEO and more like writing a sales script for a salesperson who will represent the brand in conversations the brand never sees.
The leverage is biggest at the high end. A commodity store
does not need a curated agents.md because the agent
will treat it interchangeably with the next ten stores. A brand
whose appeal depends on point of view, on edit, on staff
opinion, has nothing in the auto-generated file that captures
that. The opportunity is the same one the brand already invested
in for the human storefront: write what only you can write, and
place it on the surface that gets indexed.
The constraints to call out to clients
First, this is content. It is not a chance to negotiate the
protocol. The agent still expects the six-step transaction flow
Shopify defined in the default agents.md, and
checkout still requires explicit human approval. A studio cannot
use the template to remove guardrails or to define a one-click
agentic flow that bypasses them. What it can do is shape the
prose around the steps, surface the right entry points, and steer
the agent toward the parts of the catalogue the brand wants
featured.
Second, the rollout is theme-scoped, not store-scoped. The
templates live in the active theme and ship through theme
deploys. A multi-store merchant running different themes per
region can localise agents.md per market, which is
the right shape for an Asian portfolio where Bahasa Indonesia,
English and Japanese versions all need different framings.
Studios should plan for that in the theme architecture from day
one rather than retrofitting it.
Third, this is still emerging. Shopify has not yet folded the feature into the primary developer documentation. Field-tested behaviour comes from the changelog entry and from community write-ups. Treat the template names and fallback rules as the supported contract today, and revisit when the official docs catch up.
What we would actually change this quarter
For studios with active Shopify Plus engagements, three concrete moves.
First, audit what every client's auto-generated
/agents.md currently says. Open the URL in the
browser. Read what an LLM agent sees about the brand right now.
For most premium merchants, the gap between that text and the
brand's actual positioning will be wide enough to justify a
one-week project to close it. We would scope this the way we
scope a homepage rewrite for an
e-commerce engagement, not as
a developer ticket.
Second, treat the templates as a content deliverable, not a
technical one. The right team to write agents.md is
the same team that writes the about page and the product copy.
Engineering wires the Liquid, the brand and editorial side
decides what the prose says. Studios that hand this to a
developer end up with a slightly prettier default. Studios that
hand it to a writer end up with a sales asset.
Third, build the multi-store version up front. For merchants
with regional stores, write agents.md per market,
with locale-aware Liquid pulling currency, return windows and
shipping cut-offs from the relevant settings. This is the same
shape as a localised
product design surface,
just routed through a different file. Getting it right once
means it stays right as the catalogue grows.
The platform is signalling where the next storefront is. Premium brands that show up to that storefront with a script, instead of a default, will get the share.