Over the 2025 holidays, shoppers arriving from AI assistants converted 31% higher than any other channel (Adobe Analytics). The catch: the average retail product page is only 66% readable by the AI doing the recommending. About a third of it — sizing, materials, the answers shoppers actually want — sits locked inside images a machine can't read. So when someone asks ChatGPT what to buy, it names the store it can read. Not yours. A year ago, AI traffic converted 38% worse than paid search. This spring it converted 42% better. Same channel, opposite verdict. The stores winning it didn't outbid anyone. Their pages were just legible. I wrote up what separates a store AI recommends from one it skips — the nine things to check, and the two that close most of the gap. How many of your top product pages could an AI actually read start to finish right now?
Why don't AI assistants recommend your store? Nine times out of ten it isn't your pricing or ad spend — it's that big parts of your product pages can't be read by the AI doing the recommending. Shoppers now ask ChatGPT and Gemini what to buy; if the machine can't parse your store, you're not in the answer.
TL;DR — AI assistants recommend the stores they can read. If your product pages aren't machine-readable, you're invisible in AI shopping no matter how sharp your prices are.
- Shoppers arriving from AI assistants converted 31% higher than any other source over the 2025 holidays (Adobe Analytics).
- Yet the average U.S. retail product page is only 66% readable by AI — about a third of it is invisible to the assistants making the recommendations (Adobe Analytics).
Where did all the AI shoppers go — and why not to your store?
They went to whichever stores the AI could read. Over the 2025 holidays, traffic to U.S. retail sites from AI assistants jumped 693% year over year (Adobe Analytics), and Salesforce put AI's fingerprints on 20% of all global online holiday sales — around $262 billion (Salesforce). The shoppers are here. The only question is whether your store is legible to the software deciding where they land.
An AI referral is simple: a shopper asks an assistant what to buy and clicks a link inside the answer. If your pages read cleanly, you can be that link; if they don't, a competitor is. I spoke to a founder doing roughly $2M a year whose paid and direct traffic had flattened for two quarters while a smaller rival kept surfacing inside ChatGPT answers. His catalogue was fine. The assistant just couldn't read enough of his pages to recommend them with confidence.
Think it's your ads? Look at who's actually converting.
The reflex is to blame the ad account. But AI-referred shoppers already convert better than the channels you're paying for. Over the 2025 holidays they converted 31% higher than other sources, and revenue per visit more than doubled year over year (Adobe Analytics). Salesforce found AI-referred shoppers bought 9× more often than shoppers arriving from social (Salesforce).
So the mistake isn't the budget — it's misdiagnosing an invisibility problem as a traffic problem. More paid clicks won't put you back in the AI's answer. And these shoppers arrive further along: they've asked the assistant to narrow the field, so they reach a store closer to buying than a cold click ever is. The stores capturing them didn't outbid anyone. They were simply readable.
What does a store AI recommends have that yours doesn't?
Pages a machine can actually read. Adobe scored U.S. retail pages on how much of their content an AI can parse. The average homepage scored 75% — a quarter invisible. Product pages, where the money is, scored just 66%. And the spread is wide: the best sites hit 82.5%, the weakest 54.2% (Adobe Analytics). That 28-point gap is the difference between getting recommended and getting skipped.
The fix has a shape. Below is The 9-Point AEO Audit — AEO being whether an AI recommends your store when a shopper asks it what to buy — restated as symptoms a store owner can see.
What to check on a product page
Why it matters if it's missing
Facts in plain text, not images
AI can't quote a spec locked inside a picture
Answers to real buyer questions on the page
No text on fit, returns, or shipping = nothing to reassure a shopper
A unique written description per product
Manufacturer copy-paste reads as one of fifty identical listings
Clear product name and category labels
AI can't match you to a search if it can't tell what the thing is
Something useful loads in seconds on a phone
A slow page times out before the assistant finishes reading
Structured labels on price, stock, reviews
Unlabelled data makes the AI guess — often a rival instead
Reviews it can read as text
Ratings trapped in a widget aren't proof it can cite
A short FAQ on top products
Question-and-answer text is what assistants lift most readily
Recently updated pages
A 2023 page reads as stale to systems that favour freshness
Most stores don't fail all nine — they lose on the first two: facts trapped in images, and no plain-text answers to common questions. Fix those and you close most of the gap.
Is fixing this actually worth it? One channel already flipped.
A year ago, AI traffic looked like your worst channel: in March 2025 it converted 38% worse than paid search and email. Twelve months later it converted 42% better — a record high (Adobe Analytics). Same channel, opposite verdict — an ~80-point swing in a year.
Stores that got readable early are compounding on it now. AI-referred shoppers also spend 48% longer on site and view 13% more pages than other visitors (Adobe Analytics). This isn't a novelty channel that spikes and fades — it's the one improving fastest, and machine-readability is the entry fee.
Frequently asked questions
Do I need to be on Shopify for AI to recommend my store? No. The platform doesn't decide whether an AI recommends you — readability does. A WooCommerce store and a Shopify store both get recommended if their product pages are plain, parseable text with clear answers to common buyer questions. A slow or image-only page gets skipped on either. Fix the content, not the admin panel.
How do I know if ChatGPT can even read my product pages? Ask it. Open ChatGPT or Gemini, describe one of your best-selling products the way a shopper would, and ask it to recommend options. If your store isn't named — or the assistant gets your details wrong — parts of that page aren't reaching it. Do this for your three top sellers; the pattern usually shows up fast.
Is this the same thing as SEO? Related, but not identical. Traditional SEO gets you into Google's list of results. Getting recommended by an AI assistant depends on whether the assistant can read and trust your page well enough to name it inside an answer. A store can rank on page one of Google and still be completely invisible when a shopper asks an AI what to buy.
Will AI shoppers actually buy, or just browse? Buy. Over the 2025 holidays, shoppers arriving from AI assistants converted 31% higher than other sources and generated more than double the revenue per visit year over year, per Adobe Analytics. Salesforce found they bought 9× more often than shoppers referred from social. They arrive further along in the decision, which is why they close at a higher rate.
What's the single biggest thing hiding my products from AI? Detail locked inside images. Sizing charts, materials, and specs saved as pictures are invisible to an AI, which reads text. If your key product facts only exist inside a graphic, the assistant has nothing to quote back to a shopper — so it recommends a competitor whose facts it can read.
How fast can this change? Faster than a rebuild. Most stores recover the biggest gaps by moving product facts into plain text, answering common buyer questions on the page, and making pages load quickly on a phone. That's content work, not a replatform, and it's usually measured in weeks rather than months.
Does this matter if my traffic is mostly from ads? Yes, increasingly. AI-referred traffic to U.S. retail sites grew 393% year over year in early 2026, per Adobe Analytics, and converts better than paid channels. Building only for ads means competing on cost in a channel under pressure while ignoring the one that's growing fastest and bringing higher-intent shoppers.
The one-page action checklist
- Ask ChatGPT and Gemini to recommend your three best sellers — note whether you're named.
- Move sizing, materials, and specs out of images into on-page text.
- Answer the five questions shoppers ask most — fit, returns, shipping, materials, compatibility — in plain text.
- Confirm each product page shows something useful within seconds on a phone.
- Give every product a unique description, not a manufacturer copy-paste.
- Add a short FAQ block to your top product pages.
- Re-test in ChatGPT after two weeks and compare.
Where to start
If shoppers are asking AI what to buy and your store isn't in the answer, the problem is almost never your product — it's what the machine can read. Start with your three best sellers, get their facts into plain text, and re-test in a fortnight.
I'm Basir, co-founder of Opti Webopz. We build AI-assisted SEO, AEO, and conversion systems for clinics, agents, and ecommerce brands across the US, UK, Canada, Australia, and UAE. If you want a 20-minute teardown of your setup, the link is in my profile.