Shopify Agentic Storefronts lets your products show up, and get bought, inside AI assistants like ChatGPT and Microsoft Copilot. Since March 2026, if you run a Shopify store, your catalog is discoverable in ChatGPT by default. You didn't have to install anything.
That sounds like free distribution, and partly it is. But an AI shopping agent doesn't browse your homepage or scroll your hero banner. It reads your product data and looks at your product images, compares them with every other store selling something similar, and picks a few to show. This guide covers what Agentic Storefronts is, what the agents actually read, and how to make your products the ones they pick.
What Shopify Agentic Storefronts is
On 24 March 2026 Shopify announced that merchants can sell to ChatGPT users through Agentic Storefronts, managed centrally from the Shopify admin. Where things stand per channel:
- ChatGPT: products are discoverable by default through Shopify Catalog. Buyers check out in an in-app browser (on desktop, ChatGPT opens your store in a new tab). You stay the merchant of record, orders carry ChatGPT attribution, and there's no extra transaction fee beyond normal processing.
- Microsoft Copilot: thousands of merchants were already selling there at launch.
- Google AI Mode and the Gemini app: live for select brands so far, built on UCP (more on that below). Not default for everyone yet.
Since May 2026 it has its own page in your admin at admin.shopify.com/agentic, including recommendations to improve your product data. Find it under Sales channels → Agentic.
Can you opt out?
Partly. According to Shopify's help center, eligible products are added to Shopify Catalog automatically and you can't opt out of Catalog itself. You can block individual agentic storefronts from accessing your products. Being in the catalog also doesn't guarantee placement: each AI channel decides what it ranks and shows.
Why this is worth your attention now
Two numbers explain why every e-commerce newsletter is suddenly talking about agentic commerce:
- The traffic converts. Adobe data reported by Digital Commerce 360 shows AI-referred traffic to US retail sites grew 62% year over year in July 2026, and those visits converted 60% better than non-AI traffic, with 53% more revenue per visit.
- The plumbing is in place. ARC Report checked 11,832 stores and found 73.3% publish a UCP profile, the standard AI agents use to read a store and start checkout. Among the Shopify stores in that sample it was 98.8%. ARC is clear that its sample over-represents Shopify and that "a checkout link is not an order".
That last caveat matters. Most stores can now technically be bought from by an agent. Whether an agent actually recommends yours is a different question. Volume is still small compared with Google search, and checkout inside agents is still rough. But the shoppers who arrive this way are already further down the funnel than almost anyone else, and the stores they see are the ones whose data is easy to trust.
UCP, ACP and Shopify Catalog in one minute
You don't need to implement any of these on Shopify, but you'll see the acronyms everywhere:
| Name | Who's behind it | What it does |
|---|---|---|
| Shopify Catalog | Shopify | The structured feed of your products (titles, descriptions, images, prices, variants, availability) that AI channels read. |
| UCP (Universal Commerce Protocol) | Google, co-developed with Shopify; unveiled January 2026 | Open standard for agentic buying in Google AI Mode and Gemini. Uses existing Merchant Center feeds. |
| ACP (Agentic Commerce Protocol) | OpenAI and Stripe; September 2025 | The standard behind Instant Checkout in ChatGPT. |
The practical takeaway: every one of these pipes is fed by the same thing, your product data. There's no clever agent-only trick. The store with the cleanest titles, the most complete attributes and the clearest images wins.
What AI agents actually read from your store
Shopify's guide to optimizing products for the catalog lists the fields AI platforms consider:
- Title and description
- Images
- Product type, vendor, collections and tags
- Barcode (GTIN, UPC or ISBN)
- Variants and option names
Shopify's own summary is that AI surfaces "favor structured data, with clean attributes and real-time accuracy". In the Agentic admin you can review how your fields map into the catalog. Each field shows a coverage percentage, and title and description can come from metafields if your real copy lives there.
Why product images matter more to an agent, not less
It's tempting to think a text-based assistant doesn't care about photos. It does, for three reasons.
1. The image is required. In OpenAI's product feed spec, image_url is a required field: the main product image, showing that specific variant. Additional images are optional but supported. Google's agentic shopping runs on Merchant Center feeds, where the image rules already apply: at least 500×500 px (1500×1500 or more recommended), a clean, unobstructed view, and no watermarks, promo text or borders.
2. The image is what the shopper sees. An agent shortlists three or four products and shows them as cards. Your image is most of that card. If it's a blurry phone shot next to three crisp studio images, you lose the click no matter how good your title is.
3. The image has to match the product. When an agent recommends something, the shopper trusts the agent. If the product that arrives doesn't look like the picture, the return comes back to you, not to ChatGPT. As agents start to weigh returns and reviews, accuracy turns into a ranking factor. Marketplaces are already going this way: Etsy now requires an unedited photo of the real item, and Google requires AI-generated images to say so in their metadata.
A checklist to get your products picked
- Open the Agentic page in your admin and work through Shopify's product-data recommendations first. They're free and specific to your store.
- Fix titles. "Product type + key attribute + brand" beats clever names. An agent matching "linen duvet cover queen sage green" needs those words in the data.
- Fill in variants properly. Separate colours and sizes as real variants, each with its own image. The OpenAI feed carries a separate image per variant, so a green variant shown in blue is wrong data.
- Add barcodes where you have them. GTINs help agents match your product with reviews and other listings.
- Upload at least three images per product: a clean main shot (product filling most of the frame, no text overlays), a second angle, and an in-use or lifestyle shot. Google allows up to 10 additional images and recommends using them to show different angles or the product in use.
- Use at least 1500×1500 px for the main image. Check your sizes against our Shopify image size guide.
- Keep every image accurate. Same label text, same colour, same proportions as what you ship. If you use AI to create scenes, check each render against the real product. We counted what goes wrong most often: small text on labels and colour drift.
- Don't strip AI metadata. If an image is AI-generated and it goes into a Google feed, it must carry the IPTC
DigitalSourceTypetag. Check that your export or compression step keeps it.
Where ProductAI fits
Most of that checklist is data entry. The hard part is images: a consistent, accurate set of three to ten shots for every product and every variant, without booking a shoot for the whole catalog.
ProductAI does that from the photos you already have:
- Connect your Shopify store and pick products straight from your catalog.
- Generate a full image set per product: a clean main shot plus lifestyle and in-use scenes, in one consistent style across the catalog.
- Fidelity check on every render. Each image is compared with your original photo, and changed labels, shifted colours or missing parts are flagged before anything goes live. That's the accuracy agents and marketplaces are starting to demand.
Not sure where your images stand? Get a free visual audit of your store, or try ProductAI on a few products first.
Bottom line
Agentic Storefronts turned every Shopify store into a store that AI assistants can sell from. It didn't make every store worth recommending. Agents pick products with complete data and clear, accurate images, and the shoppers they send convert better than almost any other channel. Fix your data in the Agentic admin, then make sure every product has images an agent can trust.
