Google AI Shopping

Google rebuilt shopping around AI in two moves, six months apart. On January 11, 2026, at NRF’s Big Show, it launched the Universal Commerce Protocol (UCP): an open standard that lets AI Mode and Gemini complete a purchase without sending the shopper to a retailer’s website. On May 20, at Google Marketing Live, it turned that protocol into something merchants can actually see: Universal Cart, live with retailers including Nike, Sephora, Target, Walmart and Wayfair, alongside a new set of Merchant Center attributes built for how people ask questions in AI Mode rather than how they type into a search box.

AI Mode passed one billion monthly active users in May 2026, a year after launch, with query volume more than doubling every quarter since. The product feed that has fed Google Shopping for over a decade now has to answer a different kind of question. Here’s what changed, what’s still gated behind a waitlist, and what’s already live for Lengow merchants without any extra work. For the step-by-step configuration, see our Google AI Commerce setup guide.

What Google actually shipped in 2026

Three separate product lines now sit under the Google AI Commerce label, and they move at different speeds.

From product feed to conversational feed

A traditional product feed answers four questions: what is it, what does it cost, where do I buy it, what does it look like. Title, price, GTIN, image, brand, category, availability. That’s the vocabulary built for a shopper who clicks a search result and decides for themselves.

A conversational feed has to answer questions nobody wrote a feed spec for five years ago. Will this fit a small kitchen? Is it safe around a toddler? What does it pair with? What did other buyers say went wrong? Gemini and AI Mode don’t retrieve a match, they reason about one, and reasoning needs fields your catalogue may not have. That’s the practical meaning behind Google’s new attributes: not more keywords, more context.

Google Native Checkout: the Universal Commerce Protocol

UCP is an open standard, not a Google-only spec. It launched with Shopify, Etsy, Wayfair, Target and Walmart as co-developers, and more than twenty other companies, Stripe, Visa, Mastercard, American Express, Best Buy and Zalando among them, endorsed it on day one. It’s built to work alongside the Agent Payments Protocol (AP2), Agent2Agent (A2A) and the Model Context Protocol (MCP), so it plugs into the wider agent ecosystem rather than competing with it.

For a merchant, three things matter more than the protocol diagram:

• You stay Merchant of Record. Google Pay handles payment using saved Google Wallet details, PayPal is coming, and you keep the customer relationship, the data and the fulfilment.
• Checkout is capability-based. A merchant declares what it supports (cart, checkout, discount, fulfilment), the agent declares what it needs, and the two negotiate at the point of sale. Nobody has to implement every feature to take part.
• The functionality keeps expanding. Google added multi-item cart saving, real-time catalogue lookups and loyalty-linked pricing in April 2026, then launched Universal Cart, buy-now-pay-later through Affirm and Klarna, and two new verticals (hotel booking, food delivery) at Google Marketing Live in May.

Two caveats before this goes on a roadmap. First, Google Native Checkout is still gated: you apply through the UCP Integration Interest Form, Google approves the onboarding, and only then can you configure a feed. Second, the consumer-facing rollout is US-first. Canada and Australia are next, the UK after that. Google hasn’t published a timeline for continental Europe, so for a France-headquartered merchant this belongs on the horizon, not this quarter’s build list.

One more thing worth remembering: Google has tried native checkout before. “Buy on Google” launched in 2018 promising the same frictionless purchase inside search results, and Google quietly shut it down in 2023 after failing to pull shoppers away from Amazon and retailer sites. UCP has broader industry backing and real payments infrastructure behind it this time, which is a real difference. But the habit loop that killed the first attempt, shoppers defaulting back to Amazon or the brand’s own site, is the same one UCP has to break. Worth building for. Not worth betting an entire roadmap on.

The new discoverability attributes

Google’s Merchant Center now recognises six attributes designed for AI-driven discovery, what Google formally calls Conversational Attributes since their global rollout in May 2026. They’re optional, and Google is explicit that it doesn’t currently document a direct ranking or visibility benefit from filling them in. What it does say is that they help its models understand product nuance, variants and relationships, which is the actual currency in a world where the answer is generated, not ranked.

• question_and_answer: the FAQ your product page or support team already fields, structured as pairs Gemini can quote back.
• related_product: IDs for complementary, compatible, cross-sell or substitute items, the “what goes with this” data an agent needs to build a recommendation.
• document_link: a direct URL to a manual, spec sheet, setup guide or warranty, the kind of detail a “will this work for me” question depends on.
• item_group_title: the shared title across every variant of one parent product, so the model doesn’t treat twelve colourways as twelve separate products.
• variant_option: the specific parameter separating one variant from another, a sizing metric or a material detail, rather than a generic “size” label.
• popularity_rank: where an item sits in your own catalogue by sales or traffic, a signal beyond price and availability.

Why the timing matters

Price, location, colour, brand and availability are the five attributes Google says shoppers specify most when searching this way, according to its own year-one AI Mode report. Planning-related queries, the kind that lead to a purchase decision rather than a fact, grew 80% faster than AI Mode overall in the six months to May 2026. Recent Bain research puts the share of US consumers who now start a shopping journey on a GenAI chatbot at roughly one in eight, a category that barely existed two years ago.

None of that guarantees traffic. What it changes is the cost of being absent from the data these systems read. AI Performance Insights, the share-of-voice report Google is rolling out through 2026, will turn attribute completeness into a number a competitor can beat you on, the same way keyword rank was a number to beat a decade ago.

What’s already live in Lengow

Split this into the part that needs a decision and the part that doesn’t.

Discoverability attributes: nothing to wait for. question_and_answer, related_product, document_link, item_group_title, variant_option and popularity_rank are already native to Lengow’s Google feed structure. Map them inside your existing Google API or URL feed configuration in NetAmplify. There’s nothing new to install and nothing to wait on Google to approve.

Google Native Checkout: a short setup once you’re approved. Once Google accepts your UCP Integration Interest Form:

1. Add the dedicated Google UCP channel in your Lengow account.
2. Map your catalogue to the UCP feed structure: id as your checkout identifier if it differs from your Merchant Center product ID, plus checkout_eligibility, notice_type and notice_message for regulatory notices such as California’s Prop 65.
3. Copy the compiled feed URL Lengow generates once the configuration is saved.
4. Add that URL as a supplemental data source in Google Merchant Center, alongside your primary feed.

Talk to us about Google AI Commerce

Useful resources

• Google – Universal Commerce Protocol Guide
• Google – Introducing the Universal Commerce Protocol (January 2026)
• Google – Shopping updates from Google Marketing Live 2026
• Google Merchant Center Help – AI Performance Insights
• UCP Integration Interest Form
• Lengow Help Center – Google AI Commerce setup guide
• Lengow blog – Connecting your product feed to ChatGPT