AI Commerce: how to prepare your online store for AI shopping agents

Only a few years ago, the customer journey started with typing a phrase into Google. Today it increasingly starts with a question asked to ChatGPT, Gemini or Perplexity: “find me a light rain jacket under €150 that can arrive by Friday”. The assistant browses offers, compares them and recommends a handful of products – and in some markets it can already go through the cart and payment on its own.

For e-commerce, this is a shift on the scale of the move to mobile. Customers still buy, but more and more often an AI agent sits in between – and it “reads” your store differently than a human does. In this article we explain what AI Commerce is, what actually changed in 2026 and where to start preparing, even if agentic shopping is not yet fully available in your market.

AI Commerce is a broad term for commerce in which artificial intelligence mediates how products are discovered, compared and bought. In practice, it is worth separating two stages:

• AI-driven discovery – an AI assistant searches for and recommends products, while the purchase happens on the retailer’s website. This is already everyday reality across Europe.
• Agentic commerce – an AI agent acts on the customer’s behalf: it adds products to the cart, picks a delivery option and completes the payment within limits the customer has set.

Both stages share one thing: the decision is based on data the agent can read and trust. A store whose offer is unreadable to AI simply doesn’t make the recommendation list, no matter how good it looks to humans.

2026 showed that AI Commerce is no longer an experiment – but also that the market is still setting the rules. The key facts:

• AI traffic is growing and converting well. According to Adobe Analytics, traffic from AI tools to US retail sites grew by 393% year over year in Q1 2026. By March 2026, this traffic converted 42% better than other sources and generated 37% more revenue per visit.
• OpenAI changed course. ChatGPT’s Instant Checkout, launched in September 2025, was retired in March 2026. ChatGPT now focuses on product discovery, and checkout has moved back to retailers’ own sites. The open Agentic Commerce Protocol (ACP), built with Stripe, remains and is now used mainly to share product data.
• Google bet on an open standard. In January 2026, Google introduced the Universal Commerce Protocol (UCP), developed with Shopify, Etsy, Walmart and Target, among others, and endorsed by Adyen, Stripe, Visa and Mastercard. In May 2026, it announced Universal Cart and agentic checkout in AI Mode – first in the US, then in Canada and Australia, with the UK to follow.
• Agentic payments went live in Europe. In July 2026, Visa announced the first transactions completed by AI agents at European merchants, with more than 30 banks involved. A key element is the Trusted Agent Protocol, which helps merchants tell a verified agent apart from an ordinary bot.

What does this mean for European retailers? Google’s agentic checkout is not yet available in continental Europe, but AI-driven product discovery already is. Standards and payment infrastructure are being built right now, and stores that get their data and architecture in order early will be ready when these features reach their markets.

An agent doesn’t look at banners or animations. It analyses data and information structure. Five layers decide whether a product makes it into a recommendation:

1. Product data. Complete, consistent attributes: material, dimensions, compatibility, use case, variants. Agents answer questions like “will this fit my…”, so a vague description loses to a precise specification. Product feeds, for example in Google Merchant Center, matter too.
2. Structured data. schema.org markup (Product, Offer, ratings, return and shipping policies) helps machines read price, availability and purchase terms unambiguously.
3. Content that answers questions. Guides, comparisons and FAQs written in your customers’ language. Language models readily use content that directly addresses specific needs. Learn more about how AI understands intent in our article on how search trends are changing.
4. APIs and protocols. To check stock or place an order, an agent needs API access. New standards such as UCP (with a manifest at /.well-known/ucp) and ACP describe how a store should expose these functions. A headless architecture and well-documented APIs make this step much easier.
5. Trust and security. The store must recognise a trusted agent, handle secure payment and communicate return policies clearly. An agent that can’t find this information will pick a competitor that provides it.

You don’t need to implement new protocols right away. You’ll gain the most by getting the basics right:

1. Check how AI assistants describe your brand today. Ask ChatGPT, Gemini and Perplexity the questions your customers ask and see whether your products show up at all.
2. Audit attribute completeness in your PIM or product catalogue.
3. Add and verify schema.org structured data on product pages.
4. Keep prices, availability and shipping costs consistent across your site, feeds and structured data.
5. Describe return, warranty and delivery policies in a clear, machine-readable way.
6. Check whether your platform exposes APIs for the catalogue, stock levels and cart.
7. Review robots.txt and bot protection so you don’t block the AI assistant crawlers you care about.
8. Start tracking traffic from AI tools as a separate source in your analytics.

The first items are quick wins that also improve classic SEO and the shopping experience. The last ones require architectural decisions and are worth planning into your platform roadmap.

• Treating AI Commerce as just another plugin. Without well-structured data, no tool will improve your visibility.
• Blocking all bots. Overly aggressive protection cuts your store off from the AI assistants that bring you customers.
• Inconsistent data across channels. One price in the feed and another on the site is a signal that erodes trust – for people and machines alike.
• Waiting for a “final standard”. Protocols will keep changing, but product data, APIs and content structure will pay off whichever one wins.

Preparing for AI Commerce starts with a diagnosis. In our AI-native Commerce Audit, we check how AI systems interpret your brand and products, assess data quality, platform architecture and API availability, and finish with an action roadmap – from quick wins to longer-term development directions. Get in touch to find out what the audit would look like for your store.

If you’re also developing your content platform, see how we prepare CMS platforms for AI Search and AI agents as part of our Agentic CMS service.

Can AI agents already shop in European online stores?

AI assistants already recommend products from European stores, but checkout usually still happens on the retailer’s site. Google’s agentic checkout currently runs in the US, with further markets announced. In Europe, Visa and its partner banks are processing the first agentic transactions.

Do I need to implement UCP or ACP?

Not right away. The standards are still evolving. Start by getting your product data, structured data and APIs in order – that is the foundation for integrating with any protocol.

Will AI Commerce replace SEO?

No, it extends it. Many actions, such as structured data, content quality and page speed, help both in classic search and in AI-driven search.

How do I measure the results?

Segment traffic from AI tools in your analytics (e.g. chatgpt.com, perplexity.ai, gemini.google.com), track its conversion and revenue, and regularly check how assistants describe your products.

1. Decrypt (Adobe Analytics data) – AI Traffic to US Retailers Jumps 393% in Q1 as Agentic Shoppers Outspend Humans (2026) decrypt.co
2. Google Developers Blog – Under the Hood: Universal Commerce Protocol (UCP) (2026) developers.googleblog.com
3. Google – How we’re helping retailers thrive with new Universal Commerce Protocol features and AI tools on Google (2026) blog.google
4. Visa – Visa and banks across Europe reach the next phase of agentic commerce (2026) mynewsdesk.com
5. Exploding Topics – ChatGPT Instant Checkout is dead. Long live AI shopping (2026) explodingtopics.com