The way people find and buy products is shifting again. Not long ago, almost every online purchase began with a Google search. You typed in what you wanted, clicked through to a store, browsed, and bought.
AI tools can help customers discover and compare products. The next step varies by channel: some refer the shopper to the merchant’s checkout, while eligible direct-checkout channels can support a purchase within the AI experience. I would not treat every AI product as offering the same buying journey.
Agentic commerce extends product discovery into actions such as checking availability or helping complete a purchase. The available actions depend on the channel, store eligibility and the customer’s market.
I want to be straight about where this is and where it is going, because there is a lot of noise around it.
What Shopify has actually done
Shopify’s agentic storefronts guidance distinguishes discovery, referral and direct checkout. The store remains responsible for its catalogue and customer relationship. I check the available channel settings in Sales channels > Agentic rather than assume every store has the same rollout.
Eligibility checked on 31 October 2026. A UK-based store is not automatically excluded. For ChatGPT, Shopify currently requires sales to US customers, eligible catalogue products, completed policies and acceptance of its supplemental terms. Merchant location and customer market are different questions; check each channel separately.
For ChatGPT, customers use the merchant’s online checkout in an in-app browser or a new tab. Shopify’s current ChatGPT guidance explains that referral journey. Other supported channels may offer direct checkout when enabled and eligible. Preparing accurate data helps whichever route applies.
Why your product data suddenly matters more than your design
This is the part most coverage skips, and it is the part I find most useful commercially.
When a customer visits your store, they experience everything you have built. The photography, the layout, the copy, the trust signals, the checkout. All of that shapes whether they buy.
Product feeds, structured fields and page content can all help a system understand what is being sold. Some agents also browse the storefront, so accessible UI and visible information still matter. Accurate titles, variants, availability and policies reduce ambiguity without guaranteeing that a product will be recommended.
A title like "Classic Tote, Tan" is hard for an AI to match against a real query. "Leather tote bag in tan, structured, A4 capacity, magnetic clasp, full-grain leather" gives it something to work with. A description built entirely around the feeling of summer mornings tells the AI nothing when someone asks for a bag that fits a thirteen-inch laptop and a water bottle.
This is not an argument against good copy. Keep the copy. The point is that the functional detail has to sit alongside it. Both can be present. Most stores I see have invested heavily in how things look and far less in the underlying data. That gap is now a commercial issue.
The practical work, in order of value
If you want a store that performs well as this channel matures, here is where I would put the effort.
Start with your descriptions. Take your twenty best sellers and ask a simple question of each one. Could an AI accurately describe what this is, who it is for, and what makes it different, using only the title and description? If the answer is no, rewrite it. This improves your normal product pages at the same time, so the effort is never wasted.
Sort out your titles. Make them descriptive and literal. Product type, key material or format, and the attributes that set it apart. The AI needs to parse the title, not decode a clever brand phrase.
Tidy your variants. Inconsistent or fragmented variant data is one of the most common reasons products show up badly in AI results. Clean, consistent structure fixes that.
Complete your store policies. Shipping, returns, terms. These feed directly into how an AI answers questions about your brand, and they are good practice regardless. Wrong information about your returns policy at the point of decision is a problem you do not want, whoever's fault it technically is.
Then test what is already happening. Open ChatGPT or Gemini and search for the kind of products you sell. See what comes up, and what is missing. That is a useful spot check, but one answer does not reveal a system’s complete data sources or prove future visibility.
Your website is not going anywhere
I want to be clear about this, because the headlines tend to overstate it. For most UK brands, the traditional journey through your store will remain the main route for years. Agentic commerce adds a path. It does not remove the existing one.
What it does change is the job your product page does. A customer who arrives after researching in an AI conversation has often already decided. They are not coming to be convinced. They are coming to confirm, reassure themselves, and complete the purchase. That puts more weight on clear trust signals and a frictionless checkout, and less on the long sell.
It also changes your upsell and cross-sell thinking. A customer who buys inside an AI conversation bypasses the related-product logic built into your store. If retention and lifetime value matter to your business, and for most stores they should, your post-purchase email flows carry more weight than before. That is where the follow-on relationship now has to be built.
How AI decides who to recommend
This is worth understanding on its own, because it reaches beyond Shopify.
AI recommendations are not pulled from catalogue data alone. They draw on the same signals that have always mattered for search. Trust, authority, and reputation. Genuine reviews, credible links, a consistent presence across your website, your Google Business Profile, directories, and press. A brand that is well represented across the web gives the AI more to work with, and more reason to choose it over a competitor with a thin presence.
The useful conclusion is that the work you may already be doing for SEO and brand credibility is also building your visibility inside AI tools. It is not a separate project. It is the same fundamentals doing double duty.
A note on what comes next
I would not chase that yet. The sensible response is to get your product data clean, your brand presentation consistent, your reviews earned, and your store technically sound. Those are the right things to do regardless of how the commercial model settles. Trust, quality, and consistency tend to hold their value across whatever arrives next.
Where this leaves you
The sensible work is to check actual eligibility, clean up product information and test the route a customer can use today. I would not rebuild a store around assumptions about a future rollout.
If you want a clear view of where your product data stands today and what needs fixing first, take your twenty top sellers and run the test above this week. Be honest about what the AI could and could not understand. That short exercise will tell you more about your readiness than any amount of reading, and it will hand you a practical to-do list you can act on straight away.
For the catalogue work, start with categories, types, tags and metafields. A Shopify audit can identify the data and storefront gaps without promising AI placement.


