Shopify announced agentic storefronts last month and managed to do something quite impressive - they're making AI-native commerce actually functional and fairly simple for their platform's merchants.
No API wrestling.
No custom integrations.
No middleware.
Soon enough, (and putting AI's product selection to the side for one moment), your products will just... appear in ChatGPT, Google's AI Mode, and Microsoft Copilot. With working checkout, right there in the chat.
I've spent a while looking how this actually works, and it's worth understanding because this is the clearest picture yet of what agentic commerce looks like at scale.
What Shopify Actually Built
"Agentic storefronts" are a sales channel in Shopify. That's it.
Shopify treats AI platforms the same way they treat Facebook or Instagram - as places where people discover products and potentially buy them.
The mechanics are really straightforward too: your product data flows into Shopify Catalog (their structured product database), AI platforms query that data when someone asks about products, and if you've enabled it, customers can complete checkout without leaving the AI interface.
Two layers here that matter:
Discovery - Your products can appear in AI answers whether you've enabled checkout or not. AI platforms discover your stuff through Catalog, web crawling, or existing feeds like Google Merchant Centre.
Checkout - This is the optional bit. Turn it on for a specific AI channel (say, ChatGPT), and customers can buy directly in the chat. Turn it off, and they get sent to your site to finish the purchase.
Orders show up in your Shopify admin with channel attribution - "ChatGPT Instant Checkout" or similar - so you can track what's coming from where.
The Requirements (And Why They're Not Ridiculous)
For now, as this rolls out in its primary form, you need to be US-based, selling to US customers. Your store can't be password-protected. You need proper policies listed (terms, shipping, returns). Can't require account login before checkout.
That's the eligibility checklist. Not exactly a high bar is it?
For individual products to show up in Catalog and be purchasable through AI checkouts:
- Valid title, image, and price
- Ships to US or Canada
- Published with a working product URL
- Not flagged as mature content
- Not hidden from search engines
- Not set to "Unlisted" status
Basically, if your product would appear in Google Shopping or any other Shopify channels, it'll work here for product catalog and AI shopping.
What You Can't Do (Yet)
The AI checkout experience is simpler than your main site checkout. Some things just don't work:
Subscriptions? No.
Product bundles? No.
Local delivery and store pickup? No.
Some checkout blocks for custom fields or upsells? They might not show.
Discount codes might work, automatic discounts do work, but there's no guarantee the AI interface has a voucher field to begin with.
Google Analytics and client-side tracking pixels don't fire. You get server-to-server events only - checkout started, checkout completed. That's your lot for attribution.
This isn't Shopify being difficult. It's the reality of embedded checkouts in third-party AI interfaces. The AI platform controls the UI, not you.
How Merchants Actually Set This Up
Go to Settings → Sales channels → Agentic storefronts. (This has only just started appearing for US merchants by the way- if you're in the US and don't see it, it's on the way. If you're in the UK or elsewhere in the world, for now, sit tight.)
You'll see a list of AI platforms. Toggle each one on or off for direct selling. That's the main control.
If you want to stop a specific product being purchasable in a particular AI channel, unpublish it for that sales channel using normal Shopify product availability controls. The product can still be discovered and mentioned by AI - it just won't have a working Buy button.
Want to hide products from AI discovery entirely? Three options:
- Ask the AI platform to exclude you from their listing programmes
- Add a custom
seo.hiddenmetafield - Set product status to Unlisted
Big warning on options 2 and 3: they also hide the product from sitemaps, Google, and your own site search. This is a nuclear option. Use it when you actually want a product invisible everywhere, not just in AI.
For most merchants, the right approach is: turn on the AI channels you're comfortable with, publish the products you want sold there, and leave everything else as default.
Agentic Storefront and SEO
Here's where this gets interesting for SEO people.
Classic SEO optimised for search algorithms - you made your site attractive to Google's crawler, you structured data for their parser, you built links for their ranking system.
Agentic storefront optimisation is similar but sideways. You're making your product data attractive to AI agents whose job is to recommend and purchase on behalf of users.
Same underlying principle. Different execution.
For AI discovery, you need:
Product titles and descriptions written the way people actually describe things in conversation. "Waterproof hiking boots for wide feet" beats "Men's outdoor footwear - Model XK-9000" when someone asks an AI agent for recommendations.
Explicit attribute information. Material, fit, use cases, certifications, compatibility - all the questions a shopper would ask need clear answers in your product data. AI agents can't infer. They read what you give them.
Policy clarity. Shipping, returns, guarantees. Shopify has a Knowledge Base app specifically for structuring this information so AI can surface it accurately when someone asks about your policies.
For AI conversion, you need:
Trust signals AI agents recognise. Reviews, ratings, return policies, brand authority markers. When an agent's job is to choose products for someone, it defaults to trustworthy options. Make sure you look trustworthy in structured data, not just on your website.
Real-time accuracy on price and availability. Agents query this live. If your data's stale, you lose the sale.
Clean Catalog compliance. Every product needs proper title, image, price, and URL. Messy data means invisible products.
Agentic Commerce Optimisation
I've spoken about "Agentic Commerce Optimisation" in previous articles.
You're treating AI channels as first-class sales channels with real intent. Not just discovery surfaces. Not just brand awareness plays. Actual revenue drivers.
That means:
Calibrating which products you sell through AI checkouts. If you rely heavily on upsells or bundles, maybe those products stay web-only for now. If you sell straightforward items with standard shipping, AI checkout makes sense.
Watching where AI orders cluster. If you're getting disproportionate AI traffic on a specific product type or category, optimise those first. Better titles, better descriptions, better attribute data.
Building internal reporting that treats agentic orders as a separate cohort. Different checkout flow, different tracking characteristics, potentially different customer behaviour. Track it separately.
Aligning your entire product dataset so when AI models decide "this merchant is trustworthy for X", you pass that inference. Not just one product page. The whole catalogue.
Why This Actually Matters
Most agentic commerce announcements have been concept demos or pilot programmes with three partners. Shopify have just started making Agentic Storefront to every eligible merchant in the US as an opt-in default. That's different.
It means there's now a working model for how products flow from commerce platforms to AI agents to completed purchases. We can see what actually happens instead of speculating about what might.
The constraints are clear. The optimisation levers will soon become visible. The measurement tools exist (basic, but they exist).
If you're an SEO who's been wondering how to prepare for agentic search, this is the answer: start treating AI channels like sales channels. Optimise product data for agent readability. Build trust signals AI can parse. Track performance separately.
The skillset transfers - you're still optimising for discovery and conversion. You're just doing it for agents instead of algorithms.
What To Do This Week
If you're a US merchant and on Shopify:
Check Settings → Sales channels → Agentic storefronts. See what's turned on. Decide if that matches your risk tolerance.
Review which products are published to AI channels. Use the Products link in that same settings area.
Look at your product descriptions. Would an AI agent understand what the product is, who it's for, and why someone should buy it? If not, rewrite them.
Make sure your policies (shipping, returns, terms) are actually filled in. AI agents need this to answer customer questions.
If you're not on Shopify but you run ecommerce:
Watch this space. If Shopify made it this straightforward, other platforms will follow or merchants will migrate. The model's proven.
Start auditing your product data architecture now. Map out which fields are actually populated - titles, descriptions, attributes, variants. Check if your categorisation is consistent or if you've got three different naming conventions across different product lines.
When AI agents start querying your catalogue, they need machine-readable structure. That means proper schema markup, complete attribute data, standardised taxonomy. If your product information is buried in unstructured HTML blocks or scattered across custom fields with no consistent format, you're invisible.
Look at your trust signals. Do you have review data exposed in a format AI can parse? Is your return policy machine-readable or locked in a PDF? Can an agent verify your business legitimacy through structured data, or are you relying on visual brand cues that don't translate to API responses?
You're about to compete for ranking in AI responses. The merchants with clean, queryable catalogues and verifiable trust markers win that game. Everyone else gets filtered out before a human ever sees them.
For Shopify merchants, the jump from "I have an online store" to "AI agents can find and sell my stuff" is now just a few toggles in settings.
Getting discovered is the easy part. Getting selected is where the work starts.