Your Store Is Now Six Storefronts: A Field Guide to Shopify Agentic Commerce

A shopper asks ChatGPT for the best magnesium for sleep. It names three products, shows cards with prices and ratings, and the shopper taps one and buys - without ever opening a website. The product card in that chat was the storefront. The brand's actual site was just where the order happened to land.

That is Shopify agentic commerce, and it is not a someday thing. As of March 2026 it is on by default for eligible US stores. Your store is no longer one storefront. It is closer to six: your website, plus ChatGPT, Microsoft Copilot, Google's AI Mode, Meta, and the Shop app - every place an AI conversation can now end in a purchase.

Here is the uncomfortable part. Being eligible to show up in those channels is automatic. Being good in them is not. Most brands are switched on and badly merchandised, which is the worst of both worlds: you are in the room and losing. This is the field guide I wish I had when I started enriching stores for these channels - the six moves that actually matter, in the order I would do them, and the parts you can safely ignore.

The whole thing in one frame: discovery, then transaction

If you take one idea from this guide, take this. Selling through AI splits into two separate jobs, and almost everyone mushes them together:

1. Get discovered - does the agent find you and represent you accurately when someone asks about your category?
2. Get bought - once you are surfaced, does the agent actually sell you, and do you know it happened?

These are different problems with different fixes. You can be cited in answers and never sell anything because your Catalog is not connected. You can be perfectly set up to transact and never get surfaced because your data is thin. You need both, in that order. Every move below slots into one bucket or the other.

The demand is not theoretical, either. Shopify reported AI-driven traffic to its stores grew roughly 8x year over year in Q1 2026, with orders from AI searches up nearly 13x. Small base, steep curve. The brands that get their house in order now are compounding into that curve. Everyone else is donating impressions.

Move 1: Get discovered (be the answer)

Discovery now happens on two surfaces, and you need both.

The first is the open web that AI search reads - ChatGPT's browsing, Perplexity, Google's AI Overviews. These engines cite sources. They pull from your product pages, your feeds, your reviews, and third-party signals like Reddit. If your product pages were written for humans skimming a layout, an agent reading them for facts comes away with very little. I wrote about that specific failure in AI shopping agents are reading your product pages, and the fix - structuring pages so a machine can actually parse them - is the same fix that helps you rank without an SEO agency. Your collection pages matter here too; most have zero real content, which is a wasted, high-intent surface.

The second surface is the Shopify-native one: agentic storefronts, fed by Shopify Catalog. This is different from SEO. Instead of crawling your site, these agents read a structured product feed Shopify syndicates on your behalf. You connect once and Shopify pushes your products to ChatGPT, Copilot, Google, Meta and the Shop app - no bespoke integration per platform.

For discovery, your job on this second surface is narrow: be present and be accurate. Is Catalog connected? Are your products syncing? When you ask ChatGPT to shop your category, do you appear with the right title, price, availability and brand? Wrong price or missing variants is a discovery failure even though you are technically "there." Getting found is the whole job of Move 1. Getting the sale is Move 2 - and that line matters, because the work is different.

Move 2: Get bought (be the product the agent picks)

Here is where most of the money is, and where most brands quietly lose.

Being listed is not winning. Once an agent has a category, it ranks and recommends - and it ranks on the structured data behind your product, not on your prose. Shopify's own number is that enriched Catalog data converts roughly 2x better in AI chats than raw data. That lift is not traffic. It is the quality of your title, your attributes, your variant data, your image, your reviews. Thin or inconsistent data and the model hesitates to surface you with confidence, so it surfaces someone cleaner.

The distinction I keep coming back to is eligible versus merchandised. Eligible is automatic. Merchandised is work: a title a machine can parse ("Restore Magnesium Glycinate 200mg, 120ct" - not "Restore"), a complete attribute set, every variant with its own price and availability, a clean single-product image, reviews exposed. I broke down exactly what that enrichment requires, with a checklist, in your product catalog is now a storefront inside ChatGPT. If you only read one companion piece to this one, read that.

The honest framing: Move 2 is unglamorous data work. It is also the edge, precisely because it is unglamorous and most of your competitors will not do it.

Move 3: Pay to win the placement (when organic is not enough)

Once your data earns its place organically, there are two new paid levers - and they do not behave like the Meta playbook you already run.

Multi-channel Shop Campaigns now let you run one campaign across ChatGPT, Pinterest and Microsoft, billed on your Shopify invoice. Catalog Sponsored Products (developer preview - treat it as early) place your product inside an agent's results as a sponsored answer.

The trap is reusing your feed-ad creative here. On Meta you interrupt a scroll with a hook. In an AI surface the shopper already asked a question and the agent already decided to recommend something in your category - your job is to be the option it trusts, and the "creative" is mostly the product card. That means shorter, spec-forward copy, not "POV: you finally sleep through the night." The agent wants "glycinate form, 200mg, third-party tested," then one clean reason to pick you. Sponsor your winners - the SKUs already converting in chat - not the laggards you wish would move. And do not spend a dollar on paid placement before the underlying data is right; you would just be paying to surface a card the agent still will not confidently recommend.

Move 4: Reach the phone and inbox (the channel that fits the market)

Agentic discovery gets you found. Owned messaging - email, SMS, WhatsApp - is how you bring people back. The new channels matter, but there is one fact that quietly decides your entire mix:

WhatsApp marketing messages cannot be sent to US phone numbers. Meta does not allow it. So for a US-centric brand, SMS is your urgency channel (cart, back-in-stock, drops) and WhatsApp is your international channel (UK, EU, LATAM, India - where WhatsApp is the default app and SMS is weak). Email is every market. If a "WhatsApp for your US list" pitch lands in your inbox, that is the tell that the person selling it has not actually run it.

A few specifics worth knowing before you turn anything on: in Klaviyo, SMS, MMS and WhatsApp draw from one shared credit pool, so budget them together. WhatsApp marketing needs explicit opt-in and pre-approved message templates. SMS in the US needs TCPA consent, quiet hours, and a working STOP. None of this is hard - it is just easy to get wrong expensively. (If you want the hands-on version of running these through Claude, I tested the Klaviyo MCP end to end and nearly shipped a wrong finding, which is its own lesson.)

Move 5: Pre-test before you ship (and before you have traffic)

Most CRO advice assumes you can run a real A/B test. The math says otherwise. To detect a 20% lift on a 2% conversion rate you need roughly 19,600 visitors per variant - around 40 days for a store doing 1,000 visits a day. Most DTC stores never clear that bar on a single page, so they "test" by vibes and ship blind.

Shopify's SimGym is the interesting answer to that wall. It is an AI-shopper simulation - hundreds of simulated buyers, each with a persona, budget and intent, browsing a sandboxed copy of your store - that returns a directional A/B read in minutes, even with zero real traffic. The workflow that actually helps: the audit surfaces a theme or layout change, you build it as a draft theme, you run SimGym active-versus-draft, and you ship the likely winner - then confirm with a real test once you have the traffic.

Read it honestly, though. Shopify itself says simulated results can differ from real behavior, and it is in research preview. It is good for whole-theme and layout comparisons and for stores that cannot test at all. It is not for button-color micro-tests, pricing, or anything that hinges on real money changing hands. Treat it as a fast way to kill obviously-worse variants, not as truth. It pairs naturally with a proper conversion audit - the audit tells you what to change, SimGym pre-screens the change.

Move 6: Stay compliant on every surface

This is the move people skip, and it is the one with fines attached.

Your copy now appears on more surfaces than your website. If you sell supplements, beauty, CBD, alcohol or anything the FTC or FDA watches, the rules that govern your on-site copy - no disease claims, no fabricated stats, required disclaimers present - apply to your product cards in ChatGPT exactly the same way. There is no "it was just the AI" defense. (The deeper version of this is in the supplement marketing compliance guide.)

The genuinely useful new mechanism: Shopify's Product Disclosures field. Instead of burying warnings in description copy or theme code, you set them as structured data in the admin - Prop 65 and choking-hazard types are built in, and the FDA structure/function disclaimer, allergens or age limits go in as custom types, scoped to the jurisdictions where they apply. Because it is structured, the same disclosure renders consistently across your store, the Shop app, checkout and the AI channels. Set it once, in the right place, and it follows your product everywhere it is now sold. You still decide when a disclosure is legally required - this is not legal advice - but the mechanism finally matches the reality that your product is in six places.

The sequence: what to actually do first

Six moves is a lot. Here is the order I would run them, because order is most of the strategy:

1. Confirm you are switched on and check how you currently render. Ask ChatGPT to shop your category. Look at your own card honestly.
2. Fix the data before anything else. Enrich your top 20 SKUs - titles, attributes, variants, images, disclosures. This single step powers both discovery and transaction.
3. Get discovery right on the open web - the product-page and collection structure that helps Google also helps agents.
4. Instrument measurement. Make sure you can see AI-channel orders separately. You cannot improve a surface you cannot see, and you should not spend on it blind.
5. Only then, test paid placement - sponsor your proven winners, with spec-forward creative built for the surface.
6. Layer owned messaging and compliance throughout - the right channel per market, disclosures set in the structured field.

Data first. Paid last. Measurement before spend. None of it is exotic - it is the same operator discipline that has always separated the brands that compound from the brands that chase.

The bottom line

Agentic commerce is not a feature you toggle. It is a shift in where your store is. Your product is now being described, ranked and sold by software you do not control, in conversations you will never see, on six surfaces instead of one. The brands that win this treat their structured data as the storefront - because, increasingly, it is. The card in the chat is what the shopper sees. Make it the best-merchandised option in the room, in the order above, and the 8x curve works for you instead of against you.

If you would rather not assemble all six moves by hand, that is roughly what we built the DTC Stack to do - a skill for each move, reading from one shared brand context. But the framework above is yours either way. The agent is already recommending products in your category today. The only question is whose data it trusts enough to surface.

---

Frequently Asked Questions

What is Shopify agentic commerce?

Agentic commerce is when an AI agent helps a shopper discover, compare and buy products inside a conversation. On Shopify, agentic storefronts let customers find and purchase your products directly in AI channels like ChatGPT, Microsoft Copilot, Google AI Mode and Meta - syndicated automatically from Shopify Catalog, with checkout completed in the chat via Shop Pay.

Is Shopify agentic commerce on by default?

Yes. Agentic storefronts are active by default for eligible US stores on paid plans (since March 2026). You manage which AI channels you sell on in the Agentic section of your Shopify admin, so you can turn individual channels on or off.

Which AI channels can I sell on?

Today the agentic channels include ChatGPT, Microsoft Copilot, Google AI Mode and Gemini, and Meta, plus the Shop app. You connect once to Shopify Catalog and it syndicates your products to all of them - no per-platform integration.

Do I need the Shopify Agentic Plan?

Only if you are not on Shopify. The Agentic Plan lets businesses on other platforms (SAP, custom ERP, other carts) add products to Shopify Catalog and sell on agentic storefronts without replatforming. If you already run a Shopify store, agentic storefronts are part of your existing plan.

Does agentic commerce actually drive sales?

It is early, but the trend is real: Shopify reported AI-driven traffic to its stores grew roughly 8x year over year in Q1 2026, and orders from AI-powered searches grew nearly 13x. Enriched Catalog data converts about 2x better in AI chats than raw data. Small volume today, growing fast.

How do shoppers check out inside an AI chat?

Through the channel's Shopify-powered direct checkout. With agentic payments and stored credentials (Shop Pay), a shopper completes the purchase in the conversation - no account creation, no jumping to your site, no re-entering payment info. The Universal Commerce Protocol (UCP), co-built by Shopify and Google, is the open standard underneath.