At the end of 2025, "shopping with AI" was mostly demo videos. In 2026 it's everyday behavior: people research products in ChatGPT, Gemini, or Google's AI Mode; the agent compares options, checks stock and delivery dates, and in some markets even completes the payment. This is called agentic commerce. This article covers what actually changed in 2026, what was overhyped, and what you should do on your e-commerce site today.
What Is Agentic Commerce?
Agentic commerce is the model in which AI agents discover, compare, and buy products on a user's behalf. In classic e-commerce, the user browses category pages, applies filters, and adds items to a cart. In the agentic model, the user states a goal ("a plate-and-cup set for a 15-person kids' birthday party, delivered by Friday") and the agent matches that goal against stores' data. What the agent sees is not your design — it's your product data.
What Happened in 2026? Three Key Developments
- Google's Universal Commerce Protocol (UCP): Announced in January 2026 with partners including Shopify, Walmart, Target, Etsy, and Wayfair, this open standard defines how agents talk to store systems. In March 2026 Google added cart support, a real-time catalog capability (variants, inventory, pricing), and identity linking for loyalty programs; at Google Marketing Live in May 2026 it announced Universal Cart, which works across retailers, with checkout in a few taps via Google Pay. UCP-powered checkout is currently limited to select US merchants, with Canada, Australia, and then the UK next in line.
- OpenAI's Agentic Commerce Protocol (ACP): Launched with Stripe in September 2025, ACP made direct purchases inside ChatGPT possible (Instant Checkout).
- And an unexpected step back: In March 2026, OpenAI deprioritized Instant Checkout. Checkout is moving to merchants' own apps inside ChatGPT (Target and Instacart, among others) or straight to the merchant's website, and OpenAI said it is prioritizing product discovery. The reason: keeping constantly changing details like inventory, sales tax, pricing, and subscriptions right inside one centralized checkout proved harder than expected.
The Real Lesson: "Discover in AI, Buy on Site"
That step back is actually good news for store owners. As of 2026, the picture looks like this:
- Discovery has largely moved to AI assistants — in every market.
- Checkout still happens on your site or in your app most of the time.
So you need two things at once: data that lets an agent find and correctly understand your product, and a fast, frictionless checkout when the agent sends the shopper your way. Checkout inside ChatGPT or Google is still unavailable in most markets, but the discovery layer is already here. Stores that prepare their data today will be ready when checkout opens up.
What Do Agents Look At When Choosing a Product?
Unlike a person, an agent doesn't care about a "nice photo"; it looks at complete, consistent, machine-readable data:
| Signal | What it means to the agent | What happens if it's missing |
|---|---|---|
| Price and stock accuracy | "Can I buy it right now, and for how much?" | Products that look out of stock or mispriced get dropped |
| Variant data (size, color, dimensions) | Finding the option that fits the user's constraints | The agent can't match the right variant |
| Shipping and delivery time | "Will it arrive by Friday?" | You're invisible in delivery-constrained searches |
| Return policy | Risk assessment | The agent recommends a competitor with a clearer policy |
| GTIN / brand / model | Recognizing the product as the same item across sources | You stay an "ambiguous product" in comparisons |
| Descriptions that answer questions | Questions like "Is it dishwasher-safe?" | Marketing copy doesn't answer those questions |
A 7-Point Readiness Checklist for Your Store
- Product and Offer schema: Every product page should carry schema.org
Product+Offermarkup, with price, currency, stock status (availability), shipping (shippingDetails), and return policy (hasMerchantReturnPolicy) filled in. - Google Merchant Center feed: Keep your product feed complete, and fill in the conversational attributes Google added in 2026 for describing products in natural, conversational language.
- Real-time stock and price sync: A product that shows "in stock" in the feed but is sold out on the site erodes your credibility with agents. Update the feed when things change, not once a day.
- Question-first product descriptions: Instead of "our product is high-quality and stylish," write out dimensions, materials, care, compatibility, and use cases. Agents quote the sentence that answers the user's question.
- Open the door to bots: Make sure your robots.txt doesn't block agents such as OAI-SearchBot, PerplexityBot, or Google-Extended, and summarize your store and categories in an llms.txt file.
- Frictionless checkout: When an agent brings a shopper to your product page, you need guest checkout, express payment options, and a mobile checkout that finishes on a single screen. Don't lose a warm, agent-referred customer to a five-step form.
- Measurement: Track referrers like chatgpt.com, gemini.google.com, and perplexity.ai in a separate analytics segment, and compare that traffic's conversion rate with your site average.
Can Agents "Use" Your Site?
The next step is agents not just reading your site but using it: running a product search, adding to cart, calculating shipping. WebMCP, a new standard now being trialed in Chrome and Edge, lets your site offer agents structured "tools" for exactly these tasks. We cover it in a separate article: What Is WebMCP? Getting Your Website Ready for AI Agents.
Frequently Asked Questions
Is agentic commerce available outside the US? In-assistant checkout (buying inside ChatGPT or Google) is still limited to a small number of markets. Discovery, however, is global: shoppers everywhere research products in AI assistants and click through to stores from there.
I sell on marketplaces — does my own site still matter? Yes. Agents compare multiple sources; structured data on your own site builds brand authority and gives you a commission-free direct sales channel. A marketplace listing alone doesn't guarantee your brand shows up in AI answers.
Which platform do I need to be on to be ready? Data quality matters more than the platform. Platforms like Shopify are adding UCP and ACP integrations quickly; on custom-built sites, you need to set up the schema, feed, and API layers properly.
Conclusion
In 2026, e-commerce competition is shifting from "the most beautiful storefront" to "the most readable catalog." Where checkout will ultimately happen is still being worked out, but it's clear the winners will be the stores with complete product data, accurate stock and pricing, and a frictionless checkout. We covered choosing your stack in our guide to building an e-commerce site; if you'd like to assess your store's agentic commerce readiness together, get in touch.