Skip to main content

AI Shopping Agents Are Live — Is Your Store Ready?

· 9 min read
WebDialogAI Team

If you run a Shopify store and missed the news from last Tuesday, here's the version that matters: Ulta Beauty turned on a feature that lets Google's Gemini AI buy products from their website on behalf of customers. Not a demo. Not a beta. Live, on April 22, 2026.

No human visited Ulta's product pages. No one clicked through a search result. An AI agent identified the right products, compared options, and completed checkout — all inside Google's interface. The customer never saw Ulta's website.

That's not a far-off prediction anymore. It happened five days ago.

It's a bit disorienting to think about. You've spent years optimizing product photos, testing checkout flows, fine-tuning page titles. And now a meaningful chunk of future sales might flow through a channel where none of that matters — and what matters instead is whether your product data is structured enough for an AI to make sense of it.

The good news: this isn't just for Ulta. The Universal Commerce Protocol (UCP) that powers this was co-developed by Google and Shopify so it would work at every scale, including yours.

Here's what's actually happening, what Google changed two weeks ago that most merchants haven't noticed yet, and what to do about it this week.

What Just Went Live​

Google launched the Universal Commerce Protocol back in January with a handful of major retail partners. The idea: an open standard that lets AI agents discover, compare, and purchase products from any compliant store — inside Google Search's AI Mode, the Gemini app, and eventually other AI surfaces like ChatGPT.

On April 22, Ulta Beauty became one of the first retailers to flip the switch publicly on the full experience. A shopper can now say "find me a moisturizer with SPF 50 that ships in two days" in the Gemini app, and the AI will pull from Ulta's live product catalog, compare options, and complete checkout — without the customer ever visiting Ulta.com.

Target, Wayfair, and Walmart have been building toward the same setup since January. These are the stores that had product data clean and structured enough to plug into the protocol quickly.

If you're on Shopify, your store is already on a platform that's built to support UCP. Whether your products are actually surfaced when an AI agent goes looking is a different question, and it comes down to data quality.

What Google Changed on April 14 (That Most Merchants Haven't Fixed Yet)​

Two weeks ago, Google updated its Merchant Center product data specification — the rulebook for what product information you submit and how. A few things changed that are directly relevant to AI-driven discovery:

New attributes are now being accepted. Google added fields for product FAQs, compatible accessories, and substitute products. These aren't required yet, but they're the data that AI agents use when they're deciding whether to recommend your product over a competitor's. An AI shopping agent answering "what moisturizer works well with this foundation?" is going to favor products that explicitly list compatible items.

New shipping attributes went live. A handling cutoff time field — meaning the latest time an order placed today will still ship today — is now a supported attribute. For AI agents comparing shipping speed between products, that field matters.

Image warnings started. Google raised the minimum image resolution to 500×500 pixels across all categories. Warnings started April 14. Hard enforcement doesn't kick in until January 2027, but disapproved images mean lower visibility in AI shopping results right now. If you have older product photos uploaded before this threshold was common, some of them are probably flagged.

None of this is catastrophic. Most of it is fixable in an afternoon. But it's also the kind of thing that quietly costs you when AI agents are making product recommendations and your listing is less complete than the one next to it.

Why Product Data Is the New SEO​

For most of the past decade, ecommerce search optimization was about getting your product pages to rank. Titles, meta descriptions, schema — all designed to make pages readable by a crawler and clickable by a human.

Agentic commerce doesn't work that way. The AI agent isn't reading your page copy. It's parsing structured data: product name, description, category, variants, price, availability, images, reviews, fulfillment options. If your catalog has gaps — vague titles, missing materials, no size guides, placeholder descriptions that never got updated — the AI either skips your product or ranks it behind a competitor whose data is cleaner.

This is annoying to hear if you've spent years on copywriting and store design. But it's also genuinely fixable in a way that traditional SEO isn't. You don't need backlinks or domain authority. You need accurate, complete product information — which is something every shop of any size can do.

Four Things Worth Doing This Week​

1. Audit your 10 best-selling products for AI readability​

About 45 minutes.

Open each product in Shopify Admin and read it as if you've never heard of your own brand. Does the title clearly describe what this thing is? Does the description cover who it's for, what it's made of, what sizes or variants are available, and what problem it solves? Are all variant names human-readable ("Navy Blue, Size M" — not "Color: Option 3")?

Pay specific attention to: missing dimensions or materials, vague titles, empty tags, and descriptions that were copy-pasted from a supplier and never cleaned up. These are the fields AI agents parse when choosing between your product and the one next to it.

Fix the obvious gaps now. An hour on your top products is a reasonable use of your Sunday.

2. Connect your Shopify catalog to Google Merchant Center if you haven't already​

~30 minutes, one-time setup.

Go to the Shopify App Store, install the Google & YouTube channel, and connect your Google Merchant Center account. This syncs your product catalog automatically — it's the feed that powers Google Shopping ads, but it's also the data source for AI Mode and UCP-powered agentic checkout.

Once connected, open Merchant Center and look for "Diagnostics." Fix any disapproved items or data quality warnings. They're usually straightforward — wrong category, missing price, image resolution issue. Flagged products don't appear in AI shopping results.

If you're already connected, log in now and check the April 14 changes. There's a reasonable chance you have image warnings that weren't there a month ago.

3. Add the new AI-friendly attributes to your top products​

~1–2 hours for a product category.

Inside your Shopify product editor (or via your Merchant Center feed), start filling in the attributes Google added this month. For products where it makes sense:

  • FAQs: The top 2–3 questions customers ask before buying this product. If people regularly email you "does this work with X?" or "is this safe for sensitive skin?" — that's your FAQ content.
  • Compatible accessories or substitutes: If you sell a coffee maker and also sell filters and descaling kits that work with it, list the relationship. AI agents answering "what do I need to get started with this?" will use that information.
  • Handling cutoff time: If orders placed before 2pm ship same day, say so. Shipping speed is a real decision factor for AI recommendations.

None of this takes specialized technical skills. It's the same product information you already know — it just needs to be in the right field.

4. Send a review request to your last 90 days of buyers​

~20 minutes to set up.

Google's agentic commerce layer weighs reviews when comparing products for a recommendation. This isn't new — reviews have always mattered — but an AI agent deciding between two comparable products is going to favor the one with more recent, substantive reviews.

If you use a review app (Judge.me, Stamped, Okendo), make sure it's syncing to Google Merchant Center. Then set up a post-purchase email that goes out 7–10 days after delivery asking for a review. You probably already have one; if it's been a while since you checked the click rate, open the template and make sure the link still works.

Twenty minutes now, and it runs on its own after that.

What This Doesn't Replace​

Agentic checkout is a new discovery layer, not the whole customer journey. Once someone has ordered — whether an AI agent bought for them or they bought themselves — the relationship is still between your customer and your brand.

That's where the experience gets personal: "Did this ship yet?" "Can I swap the color?" "This didn't fit — what's the return process?" Those questions happen after the purchase, on your website, via chat or email. A well-trained AI layer on your website handles the routine ones automatically and routes the real issues to a human — so the experience holds up across the whole order, not just the moment the AI picked your product.

Hang in there. See you tomorrow.

WebDialogAI gives your website an AI chat with seamless human handover — answers customer questions instantly, escalates to a real agent when it matters. Get started free or install on Shopify.

Sources: