ChatGPT sent 20.5 million visits to US online retailers in Q1 2026 alone, and 72% of AI-shopping users now treat these tools as their primary way to research a purchase. What decides whether they recommend you isn't ranking — it's whether your product pages are actually readable by an AI.
Signal
Product research inside an AI shopping assistant (ChatGPT, Gemini, etc.)
Context
Plain-language product facts, structured data, and recent reviews, captured and structured
Response
Audit top product pages against what an AI assistant needs to recommend them
Owner
E-commerce and product marketing lead
ChatGPT sent 20.5 million visits to US online retailers in the first quarter of 2026 alone. Not from ads. Not from search rankings in the way most retailers still think about them. From shoppers typing a question — "what's the best running shoe for flat feet," "which air fryer is easiest to clean" — into an AI assistant and getting back a direct answer, often with a product named and a link attached.
This isn't a future trend to prepare for. It's already 72% of AI-shopping users treating these tools as their primary product research method, ahead of a plain Google search, ahead of Amazon's own search bar, ahead of asking a friend. And 77% of them say it makes the whole decision faster — which means less time for a retailer to catch their attention once they've started forming an opinion elsewhere.
Why this is different from SEO
Traditional SEO is built around a ranking position on a page a human scans — ten blue links, a few ads, a map pack. You optimise to be near the top, and a human does the comparing. AI shopping assistants don't work that way. They synthesise a single answer, sometimes naming two or three options, and the shopper usually acts on that answer directly. Ranking eighth on Google doesn't matter if the AI never mentions your product by name at all — there's no page two to fall back on. What matters instead is whether the AI can extract clean, confident facts about your product: what it costs, whether it's in stock, what it's actually for, and who it isn't a good fit for.
Five things AI shopping assistants need from a product page
AI shopping assistants synthesise a single answer instead of showing a ranked list of links, so a product page needs to make these facts impossible to miss:
- A plain-language answer up front. "What is this for, who is it for" needs to be stated near the top of the page, not buried three paragraphs into a marketing description.
- Structured data. Product, Offer and AggregateRating schema make price, availability and rating machine-readable rather than just visually obvious to a human eye. Without it, an AI is often guessing.
- Real, visible, recent reviews. AI tools weight review volume and recency heavily when deciding what to recommend — a page with reviews from 2022 carries less weight than one with reviews from last month.
- Clear differentiation between your own similar products. If three of your product pages all sound the same, the AI gets confused about which one to recommend — and an AI that's unsure tends to recommend a competitor instead.
- Comparison-ready content. Shoppers ask AI tools "X vs Y" constantly. A product page that states honest trade-offs — what it's better at, what it's not — gets cited more often than one that only sells.
The measurement problem
Here's the uncomfortable part: most e-commerce platforms have no built-in visibility into AI-referral traffic, the way they've had for Google and Meta ads for over a decade. If a shopper arrives from chat.openai.com or gemini.google.com, it often shows up as "direct" traffic or gets missed entirely in the usual reports. The practical fix isn't waiting for perfect attribution — it's checking referral sources for these domains now, treating any uptick as an early signal, and starting to fix product pages before the reporting catches up.
Where to start
Pick your ten highest-traffic product pages. For each one, ask: could an AI answer "what is this, who's it for, how much does it cost, and is it actually good" using only the information on that page — without guessing, without inferring, without needing to visit a competitor's page to fill in the gaps? If the answer is no, that page is invisible to a growing share of your customers before they ever see it.
Sources: Capital One Shopping Research, 2026
Want a check on whether your product pages would pass that test? Book a working session and we'll run it against your highest-traffic listings.