Agentic commerce for store owners: what changed in a year, and what to do now

If you run an online store, someone has probably told you that AI assistants are about to do your customers' shopping. Part of that is true. Agentic commerce, where an assistant such as ChatGPT, Gemini or Copilot finds products, compares them and sometimes completes the purchase, went from announcement to working product between September 2025 and October 2026. It also had its first public retreat.

This guide is for store owners on Shopify, WooCommerce or a custom site, and for marketplace sellers, who want to know what actually changed, what to do this quarter and what to leave alone. This area changes monthly, so every claim is dated, and the guide is current as of 15 October 2026.

I build automation for Shopify stores and marketplace sellers, and AI agents for businesses. I do not run a store, so this is a builder's view: what the platforms shipped and the work it creates for you. If you are new to automating a store, start with my guide to e-commerce automation.

What agentic commerce means for a store

There are three layers.

1. Discovery. The assistant answers "find me a waterproof walking jacket under £150" from product data it already holds: feeds, catalogues and your product pages. If your data is thin or wrong, you are not in the answer.
2. Checkout. Either the assistant sends the shopper to your store, or it completes the purchase inside the chat through a checkout your platform or payment provider exposes.
3. Payment and trust. Protocols that let an agent prove the shopper approved this cart at this price, so payment providers and merchants can accept the order.

In the setups launched so far, you stay the seller of record. You take the order, ship it and handle returns and support. The assistant is a new shop window that you do not design.

What changed from September 2025 to October 2026

When	What happened	What it means for you
Sept 2025	OpenAI launched Instant Checkout in ChatGPT on the Agentic Commerce Protocol (ACP), built with Stripe, starting with US Etsy sellers	Buying inside a chat became real
Nov 2025	Perplexity and PayPal launched Instant Buy for US shoppers	More assistants, same pattern
Jan 2026	Google and Shopify announced the Universal Commerce Protocol (UCP); Microsoft launched Copilot Checkout in the US, using ACP	Two protocols, each backed by large platforms
Mar 2026	OpenAI scaled back Instant Checkout; Shopify's Agentic Storefronts reached ChatGPT	Discovery matters more than the buy button
Oct 2026	Google began rolling out Universal Cart in the US	One cart that spans several shops

Google's January 2026 UCP announcement names Shopify, Etsy, Wayfair, Target and Walmart as co-developers. It says retailers remain the seller of record, that shoppers can pay with Google Pay on eligible listings in AI Mode and the Gemini app, and that UCP works alongside Google's Agent Payments Protocol (AP2) and the Model Context Protocol.

On 15 October 2026 Google started rolling out Universal Cart across Google Search and the Gemini app in the US. It keeps items from different retailers in one cart, alerts the shopper to price drops and restocks, and either checks out through UCP or moves the items to the merchant's own cart.

On Shopify, Agentic Storefronts put your products into ChatGPT, Google AI Mode and Gemini, Microsoft Copilot and Meta's AI surfaces. Shopify's help centre says the feature is "active by default for eligible stores". Etsy's help centre, meanwhile, explains how US buyers can buy some Etsy listings directly in ChatGPT, Gemini, Google AI Mode and Copilot, as guest checkouts.

The lesson from OpenAI's pull-back

In March 2026 OpenAI told TechCrunch that "the initial version of Instant Checkout did not offer the level of flexibility that we aspire to provide". ChatGPT now concentrates on product discovery, and shoppers mostly finish the purchase on the merchant's own site. Checkout through apps inside ChatGPT is still possible for now.

Three things follow:

• The buy button was the least durable part. It lasted about six months in its original form.
• Discovery survived. Assistants still need accurate, complete product data before they can recommend you.
• Do not bet on one protocol. ACP, UCP and AP2 overlap, and some may merge or fade. Prefer work that serves all of them: product data, stock and policies.

How to get your store ready for agentic commerce

Most readiness work is unglamorous, and none of it is wasted if the protocols change again.

1. Fix product data first. Titles that say what the item is. Complete attributes: size, material, colour, dimensions, compatibility. Barcodes (GTINs) where you have them. Variants as real variants, not separate products. An assistant answering "will this fit a 2019 Golf" needs that fact written down somewhere it can read. For large catalogues, an AI listing pipeline can draft missing attributes for a person to check.
2. Keep feeds complete and fresh. Google's merchant guide for UCP builds on your existing Merchant Center feeds, so that feed is where most stores should start. OpenAI accepts product feeds from approved partners (applications go through chatgpt.com/merchants), and its feed documentation recommends a full feed once a day plus updates during the day. On Shopify, check what the platform already sends before building anything.
3. Mark up product pages. Add Product structured data with price, availability, shipping and returns, and make it match the feed. Google's documentation treats pages where people can buy as "merchant listings", which support shipping and return details.
4. Write policies an agent can quote. Returns window, who pays return postage, delivery times per country, warranty. One clear page each, in plain sentences, with the same numbers you give Merchant Center. In January 2026 Google also announced new Merchant Center attributes for answers to common product questions, compatible accessories and substitutes.
5. Make stock accurate everywhere. An agent that sells something you do not have creates a cancellation in a channel you do not control. If you sell on several channels, multichannel inventory sync comes before any AI work.
6. Let the right crawlers in. OpenAI's crawler documentation says sites that opt out of OAI-SearchBot "will not be shown in ChatGPT search answers". Check robots.txt and your CDN's bot settings, which can block crawlers you never meant to block.
7. Measure it. Set up an analytics segment for visits referred by AI assistants, so you can see whether this channel matters for your store before you spend more on it.

What to ignore for now

Some of the noise is safe to skip this year:

• Building your own ACP or UCP integration on Shopify. The platform already does it. On a custom stack, join Google's UCP waitlist or apply to OpenAI's merchant programme when your catalogue and volume justify the work.
• "Rank in ChatGPT" packages. Nobody can promise placement in an assistant's answer. The levers you control are the ones in the checklist above.
• Experimental agent payment rails, such as stablecoin payments for agents. Worth watching, not worth building for at a small or mid-size store today.
• A shopping chatbot added because agents are fashionable. It earns its place only by answering real questions from your own data.
• Rewriting every description "for AI". Fill in missing facts first. Tone comes later.

The five-year view

McKinsey estimates that up to $1 trillion of US B2C retail revenue could be orchestrated by agents by 2030, and $3 trillion to $5 trillion globally. Read "up to" carefully: it is a ceiling under a broad definition, not a promise that agents will press "buy" on that much.

The demand signal today is growth, not size. Adobe data reported by TechCrunch shows AI traffic to US retail sites up 393% year on year in Q1 2026, and AI-referred visitors converting 42% better than other traffic by March 2026. For most stores, assistants are still a small source of visits.

My reading: the protocol names will change, and the inputs will not. Product facts, prices, stock, delivery promises and return terms are what every assistant reads. A store with those in order in 2026 will be ready for whichever system wins by 2031, and the same clean data also improves your own site search, marketplace listings and ads.

Build, buy, or let the platform do it

• A Shopify store selling on one channel. Mostly let the platform do it. Check the Agentic settings, fill in product data, tighten policies.
• Several channels or a custom store. Keep one master catalogue and push it to Merchant Center, any AI feed you are accepted for and your marketplaces, with stock kept in sync. That is automation work. My marketplace and e-commerce plans on the pricing page start from $2,500 for an MVP (one marketplace connected through its official API, plus one automation such as listing updates), and from $5,500 for Flexible (two or more channels, stock and orders kept in sync, a dashboard). These are starting prices, and every project gets a fixed quote after a short brief. The e-commerce automation page shows the kind of work involved.
• Agents of your own. An assistant that answers product questions on your site, or a connection that lets Claude or ChatGPT read your store data. That is custom AI agent work, quoted from a brief.

FAQ

Do I need to do anything for agentic commerce if my store is on Shopify?

Shopify says Agentic Storefronts are active by default for eligible stores, so your products may already be discoverable in ChatGPT, Copilot, Google AI Mode and Gemini. Open Sales channels, then Agentic, in your admin to see which channels are switched on. The work that remains is yours: complete product attributes, accurate stock and clear shipping and return policies. Those decide whether an assistant recommends you.

Can I still sell through ChatGPT after OpenAI scaled back Instant Checkout?

Yes, in a different shape. Since March 2026 ChatGPT has focused on product discovery, and shoppers usually complete the purchase on your own store's checkout. Shopify merchants reach ChatGPT through Agentic Storefronts. Other merchants can apply to OpenAI's merchant programme with a product feed, which OpenAI currently onboards for approved partners. Treat it as a discovery channel and measure it like one.

Will AI shopping agents replace my online store?

Not in the setups launched so far. Google and Microsoft both say the merchant stays the seller of record: you take the order, ship it and handle returns and support. What changes is where the shopper decides. More decisions may happen inside an assistant, so your product data, prices and policies need to be readable there, while your store keeps the fulfilment and the customer relationship.

If you want help getting a store ready for AI channels, send me a short brief: which platforms you sell on, roughly how many products, where your product data lives today and what you want working first. You can start with the marketplace MVP plan, read about AI agents for business, or see how to connect Claude and ChatGPT to your own data in MCP for e-commerce.