TL;DR: Shopify runs the shops behind a large slice of the internet’s online retail, so it can see what happens when an AI chatbot sends someone to a store. On 11 August it published those numbers. Visits arriving from AI assistants roughly tripled over the year; visits from ordinary search grew 12%, from a base so much bigger that ordinary search still sends most of the traffic. The finding worth your attention is not the growth. It is that shoppers bought twice as often when the AI had read a clean, machine-readable copy of the product facts rather than reading the page and guessing.
First, three terms, in plain words
A session is one visit to a site by one person. An AI-referred session is a visit that started inside a chatbot: someone asked ChatGPT or Gemini or Perplexity about a product, the answer mentioned a shop, they clicked through.
A product detail page is the page for one single item: the one with the price, the size chart and the buy button.
Structured data is the part people skip, and it’s the point of the whole story. When a machine wants to know what a product costs, it has two options. It can read the page the way you do and work it out from the text and the layout, which is called scraping and involves a lot of guessing: is “£49” the price, the old price, or the delivery threshold? Or it can be handed a labelled list, where the price sits in a field called price and the stock status in a field called availability. That labelled list is structured data. Shopify keeps one for the products in its shops, calls it Catalog, and says it covers more than a billion items.
The numbers
The figures come from Shopify’s own second-quarter commerce data, published on 11 August and reported by Search Engine Land two days later.
AI-referred sessions to Shopify storefronts grew 197% year over year. Orders from those sessions roughly tripled. Organic search — the free listings in Google and everywhere else — grew 12%.
Read those two numbers the wrong way and you conclude search is finished. Search Engine Land’s headline puts the correction in the right place: organic still leads traffic. A 12% rise on a very large number is more actual visits than a 197% rise on a small one.
There’s a second thing hiding in the growth rate. In Shopify’s first-quarter data, AI-referred sessions were up more than 8x year over year and orders nearly 13x. A quarter later that’s 197% and roughly 3x. Still fast, clearly slowing. Anyone budgeting on Q1 multiples now has a data point against that.
One more figure from the same dataset: about half of AI-referred sessions land directly on a product detail page, rather than on a homepage or a category listing.
AI does not win every category by the same margin
Shopify split conversion rates by product category, and AI-referred shoppers converted better in nearly all of them — but the size of the gap moves a lot. Watches: 2.4x better than organic. Necklaces: 2.3x. Apparel: 1.6x.
The pattern behind that is easy to hold in your head. If a purchase turns on comparing specifications — case diameter, water resistance, chain length, compatibility with the thing you already own — an AI assistant is genuinely good at the work, and the person clicking through has already done the comparing. If a purchase turns on whether you like how it looks, the assistant can’t do much for you, and you go back to browsing. Shopify also found AI introduces first-time customers at about 1.3x the rate of organic search.
So the honest position for a client conversation is neither “AI is taking your traffic” nor “AI is a rounding error”. AI referrals are still a small share of visits, they are growing, they buy at a good rate, and how much they are worth depends heavily on what the client sells.
The structured-data finding
Here is the sentence to take away, in Shopify’s own framing: when AI search used structured Catalog data to find and recommend products, the shoppers it referred converted at 2x the rate of shoppers from AI sessions that relied on scraped or third-party product feeds.
There’s no mystery in the mechanism. An assistant working from scraped text gets things subtly wrong: a stale price, a variant that’s out of stock, a spec pulled from the wrong row of a table. The shopper arrives, finds the page doesn’t match what they were told, and leaves. Bad product data doesn’t stop the referral happening. It kills the sale after the click.
Most merchants aren’t on Shopify Catalog, but the principle isn’t Shopify’s. It’s whether the machine-readable version of your product page agrees with the visible one. That’s schema markup — the labelled product facts you can embed in any page — and it’s whether your prices, stock status and specs in that markup are actually current.
What to check this week
- Does every product page carry product markup with price, currency, availability and SKU, and does it match what the page displays right now? Stale markup is worse than none.
- Do the specs a buyer would compare on actually appear as text on the page? Measurements trapped in an image are invisible to everything.
- Look at your own category split. If a chunk of your catalogue is spec-led, that’s where AI referral work pays back first.
- Check where AI-referred visitors land. If they arrive on product pages, those pages carry the whole visit, and a thin one has nowhere to recover.
Where this gets expensive
Auditing one product page against its own markup takes a couple of minutes. Doing it across four thousand SKUs, then again after the next theme update quietly drops a field, is work no retainer has room for. Preferium runs 47 automated checks over every page on a site, scores it 0–1000, and then fixes what it finds, ships the change and re-checks the live page in a real browser. Broken or missing structured data, and markup that has drifted away from what the page actually says, sit squarely in that set. More on how the system works.
Related, and from the same week: how ChatGPT builds the short snippet it stores about your page. The same logic applies to a price.
Key takeaways
- Shopify’s Q2 data, published 11 August: AI-referred sessions +197% year over year, organic search +12%, AI-referred orders roughly tripled.
- Organic search still sends far more traffic. Both channels grew; neither replaced the other.
- Growth in AI referrals is decelerating — Q1 was 8x sessions and nearly 13x orders, Q2 is about 3x.
- Shoppers referred by AI reading structured catalog data converted at 2x the rate of those referred by AI working from scraped or third-party feeds.
- The conversion advantage concentrates in spec-led categories: watches 2.4x, necklaces 2.3x, apparel 1.6x. AI also brings first-time customers at about 1.3x the rate of organic.
- Practical version: keep your product markup accurate and current, and put comparable specs in text rather than images.
