Our man on the ground, Nick Young, was in Toronto for DotDev, Shopify's developer conference. Two days of sessions, slide photos and notes, which he turned into a proper field guide for the team.

I've been through all of it. There's a lot to unpack, but one thing matters more than the rest, and it's not a new feature.

Shopify is positioning itself as the layer that AI assistants shop through.

Not a marketplace. Not a search engine. The plumbing underneath: live product data, pricing, availability and checkout, exposed directly to whatever assistant the customer is using. When someone asks ChatGPT or Gemini or whatever comes next to find them a product, Shopify wants the answer built from its data, with a basket ready at the end.

Now, the obvious question: is anyone actually shopping this way?

Shopify's own Q1 numbers say yes, and faster than I expected. AI-chatbot referral sessions to Shopify stores grew more than 8 times year on year. Orders referred by AI grew nearly 13 times. And here's the bit that should get your attention: visitors arriving from AI referrals converted 49% higher than organic search, with a 14% higher average order value.

Keep in mind, that's growth from a low base. 13 times a small number is still a smallish number. But the direction is not in doubt, and the conversion gap tells you something important. By the time an AI assistant sends someone to your store, the research is done. They're not browsing. They're buying.

Which brings me to the part most merchants are going to get wrong.

When an assistant assembles an answer, it doesn't rank ten blue links and let the customer decide. It filters. It looks for products that match the full question: the size, the material, the compatibility, the delivery window, the return policy. If your product data doesn't contain the attribute the customer asked about, you are not ranked lower.

You're not in the answer at all.

That's a completely different problem to the one SEO has spent twenty years solving. An incomplete product page used to cost you a position. Now it costs you the entire consideration set.

This is something we've been banging on about for a while. If your returns policy, shipping information or terms are blocked from crawlers, or simply don't exist on the site, you're invisible to the agents doing research on your customers' behalf. Most agencies won't flag this. SEOs won't, ads specialists won't, UX teams won't. It only shows up when you specifically look at the site through the eyes of a machine trying to complete a purchase for someone.

The fix is not another dashboard. It's not a tool that counts how many times a chatbot mentions your brand and hands you a graph. It looks impressive; it's also useless.

The fix is boring, and that's good news, because boring is controllable:

• Product attributes actually filled in. Not marketing copy. The smallest complete set of facts needed to answer real buying questions: what it is, who it's for, what it works with, what's included, and what would make it the wrong choice.
• Policies visible and crawlable. Shipping, returns, warranty. In the HTML, not rendered by JavaScript after the fact.
• Consistent structured data, so the facts on the page agree with the facts in your feeds.
• Real customer language in reviews and FAQs, because that's the language the questions arrive in.

None of this is glamorous. All of it compounds.

One more thing from Nick's notes, because it's the piece most coverage will skip. Shopify is also building the loop that connects a recommendation to a measured outcome: targets, annotations, queryable data inside the admin. In plain English, they want the merchant to see what was suggested, what was changed, and what happened afterwards. That's how we've run client work for years. It's reassuring to see the platform catching up, because it means the tools will start rewarding the businesses that operate this way and exposing the ones that don't.

So here's my suggestion. This week, pick your ten best-selling products and ask a simple question: could an AI assistant answer the five most common pre-purchase questions about each one, using only what's on the page and in your product data?

If the answer is no, you don't have a rankings problem. You have a data problem, and it's getting more expensive every quarter.

Full disclosure: this is exactly why we built SKAW. It's our Shopify app that audits your store the way an AI agent sees it: missing attributes, weak descriptions, incomplete policies, blocked bots. You get a readiness score with the fixes ranked by impact, and it tracks which AI agents are already visiting your store and which products they're looking at. It's free, and if the ten-product test sounds like homework, SKAW does the homework for you.

Reply and tell me what you find. Genuinely curious how many stores pass their own ten-product test.

Thanks to Nick for the miles and the notes. Normal video service resumes when the project pile allows.