On May 28, 2026, Shopify quietly handed merchants something the AEO crowd has wanted for a while: theme-level control over the files AI agents read from your store. You can now add Liquid templates — templates/agents.md.liquid, templates/llms.txt.liquid, and templates/llms-full.txt.liquid — under Online Store > Themes > Edit code, and the system falls back to the agents.md template, then to Shopify's auto-generated default.
In plain terms: you now get to influence what ChatGPT, Gemini, and other AI agents say about your store instead of leaving it entirely to their interpretation. For brands fighting to be the one an agent recommends, this is implementation-level answer-engine optimization, finally with a knob you control.
What changed on May 28
Until now, AI agents that crawled your Shopify store read whatever Shopify auto-generated plus your public pages. You had limited say over the framing. The May 28 release adds three editable Liquid templates:
templates/llms.txt.liquid— a concise, structured map of your store for AI to read.templates/llms-full.txt.liquid— an expanded version with more detail.templates/agents.md.liquid— guidance aimed at how an agent should interact with and represent your store.
The fallback order matters: the system uses your templates where you provide them, then falls back to the agents.md template, then to Shopify's auto-generated default. So you're never starting from zero. You override only what you want to control and let Shopify handle the rest.
Because these are Liquid templates, you can pull in live store data — collections, featured products, policies — rather than hardcoding a static file that goes stale the moment you change your catalog.
Why this matters for discovery
AI assistants are becoming a real discovery surface, a shift we track across posts like how to get recommended by ChatGPT and the Claude AI optimization guide. When a shopper asks an assistant "what's a good store for X," the assistant reads available signals and decides how to describe and rank you.
If you leave that to the default, you're trusting the agent's guess. With these templates, you can:
- Lead with what makes you different instead of a generic product dump.
- Prioritize your best collections so agents surface what actually converts.
- State the facts you want quoted — shipping terms, guarantees, materials, sourcing — so the agent's answer is accurate and on-brand.
How it interacts with your product data
A common misread: these files do not replace your product catalog. Shopify's structured product data still feeds prices, availability, and attributes to agents. The agents.md and llms.txt templates sit alongside that data and add context and framing.
So your job splits into two layers. Keep the structured catalog clean and accurate, the foundation we stress in the Shopify checkout conversion leak audit and across our agentic-commerce coverage. Then use the new templates to add the narrative — who you are, what to highlight, how to describe the brand — that raw catalog fields can't carry.
What to put in these files
Keep it tight and true. Agents reward clarity, not keyword stuffing.
For llms.txt, a clean structured map:
- A one-line description of what you sell and who it's for.
- Your top collections and flagship products, linked.
- Key policies: shipping, returns, guarantees.
- Links to your most useful content — buyer guides, comparisons, sizing.
For llms-full.txt, expand the above with more detail per item: the why behind a flagship product, the materials, the differentiators.
For agents.md, guidance and framing:
- How you want the brand described in one or two sentences.
- What to emphasize (e.g., sustainability, fast shipping, a niche specialty).
- Any accuracy guardrails — facts you want stated correctly.
Resist the urge to oversell. AI assistants weight trustworthy, specific, consistent information. Honest and precise beats hypey every time.
Where this fits with paid visibility
Be clear-eyed about what AEO does and doesn't do. Optimizing these files improves how agents represent you, but AI discovery is earned and unpredictable. You don't control when or whether an assistant surfaces your store, and you can't dial it up the way you can dial up ad spend.
That's exactly why this complements, not replaces, paid visibility. Paid ads are bought and controllable; AI recommendations are earned and shaped. The smart play is both: clean agent files to win the recommendations you can earn, and paid campaigns to guarantee the reach you can't. Once AI traffic starts arriving, the next job is converting AI shopping traffic once it lands.
What to do this week
- Go to Online Store > Themes > Edit code and add the three Liquid templates.
- Start with
llms.txt— a clean, structured map of your store. Keep it short. - Write a tight
agents.mdwith one or two sentences of brand framing and your accuracy guardrails. - Use Liquid to pull live data for collections and featured products so the files don't go stale.
- Test how agents describe you before and after, then refine. Treat it like an iterative SEO file, not a one-time setup.
If you're not on Shopify yet and want this level of native AI control, you can start a Shopify trial and configure your agent files from the start. For the strategy layer behind these files, see how to get recommended by ChatGPT.
You finally get a say in what AI tells shoppers about your store. The merchants who fill these files with clear, honest, specific signals will be the ones agents describe accurately and recommend confidently. The ones who leave the default will keep wondering why the assistant got their brand slightly wrong.