Ask Google for the best carry-on suitcase, and you’ll get 10 links, a handful of ads, and a long evening ahead of you.
Ask ChatGPT the same question, and you get three suitcases, each with a neat little reason attached: one’s the lightest, one can survive being hurled around by baggage handlers, and one fits under the seat on a budget airline.
It’s quick, it’s tidy, and it raises a slightly uncomfortable question for brands: Who decided those three suitcases were the best?
Plenty of shoppers are already acting on the answer. According to Salsify’s “2026 Consumer Research” report, 22% of shoppers bought a product in the past year because an AI search tool or shopping assistant recommended it, up five points from 2025.
This is where generative engine optimization (GEO) comes in: the work of making sure your products are the ones AI pulls into its answer. But before you can optimize for “the best,” you need to know what an AI actually means when it says “best.”
Nearly one in five shoppers (19%) globally discover new products and brands on tools like ChatGPT and Gemini, and in the U.S., it’s one in four. On the research side, 22% of shoppers use AI search tools, putting them ahead of product review websites (19%) and online forums like Reddit (14%).
And that’s just the people who call it a primary channel. Salsify’s “Ecommerce Pulse Report: Q4 2025” found that 64% of shoppers have used AI shopping tools to discover or research new products at some point.
Meanwhile, as AI product discovery grows, the channels that owned discovery a year ago are losing their grip. Purchases driven by influencer recommendations dropped 16 points to 23%, and purchases of viral products fell 17 points to 17%.
Younger shoppers are leading the charge: 32% of millennials and 31% of Gen Zers have bought something on an AI recommendation, compared to just 3% of baby boomers.
There’s a catch, though. Only 14% of shoppers say they trust AI shopping tools and use them regularly. Another 27% trust them for some purchases but verify with other sources, and a third (33%) don’t use them at all.
Salsify calls this the AI trust gap, and it’s a big part of what’s rewriting the AI path to purchase. Shoppers are happy to ask AI for the best option… they’re just not ready to take its word for it.
Old-school search asked one question: Does this page contain the right words?
A large language model (LLM) asks something much harder: Does this product fit what this person needs, and can I back that up?
AI doesn't play favorites here. It can only recommend what it can back up, which usually comes down to three things: deep product data, customer sentiment and reviews, and verified attributes.
When Salsify asked shoppers what makes them trust an AI recommendation enough to buy, detailed product descriptions and specifications came out on top (31%).
Specs alone won’t cut it, though. Nobody really asks for “the best suitcase.” They ask for the best suitcase for a two-week trip with a toddler. AI shopping tools favor context, so your content needs to explain who a product is for and why, as well as important info like how much it weighs.
A quarter of shoppers (25%) say customer reviews or ratings make them trust an AI pick. Flip that around, and 42% have abandoned a purchase because of no, low, or negative reviews.
AI reads those reviews too. If your product page says “all-day battery” and a couple hundred customers say it’s dead by lunch, guess whose version makes it into the answer? If you said the customers’, you’re absolutely right.
When shoppers describe what makes a product high quality and high value, they list things you can prove, like:
Notice how none of those are opinions. They’re facts that are either evident in your product data or not (which is why product data accuracy matters so much).
Shoppers also trust an AI recommendation more when it explains why the product suits them (25%) or shows how it compares with similar products (21%). AI can't do either without detailed product data to draw on.
Your product page is just one of the places AI looks. It also checks retailer listings, reviews, forums, and third-party write-ups to see whether they all tell the same story.
Funnily enough, that’s exactly what shoppers do afterward. Before a big-ticket purchase, 30% of shoppers review four to six channels, and 11% review seven to 10. In the U.S., more than half (56%) check four or more.
So your product story gets audited twice: once by the machine, and once by the human double-checking the machine.
This is what digital shelf authority comes down to. Being the loudest brand on one channel won’t get you there. Telling the same, accurate story on every channel an AI or a shopper might check just might.
And product data accuracy is where plenty of brands come unstuck:
A shopper who spots a mismatch gets suspicious. An AI that spots a mismatch gets cautious, and when your PDP and the shoppers’ consensus disagree, AI will have no qualms about dropping your product from the shortlist.
The good news is you don’t need one content strategy for robots and another for people.
Salsify’s research found that the same information AI feeds on (accurate product data, useful context, and authentic conversations) is what makes shoppers trust a recommendation enough to complete a purchase.
In practice, the report points to three GEO priorities:
Think of it this way: Structure makes your product details easy for AI to find and quote, and enrichment gives it a reason to choose you over a competitor.
Take a set of bed sheets. "Luxuriously soft" is lovely, but an AI can't do anything with it. "100% organic cotton, 300 thread count" is a fact it can quote. Add "stays cool for hot sleepers," and now it knows who to recommend them to.
If you want the full story on what happens when that data’s missing, here’s why AI agents hate incomplete product content.
One caveat: Brand recognition is still the second-biggest trust trigger at 27%. GEO doesn’t replace brand building, but it does make sure your brand shows up when it counts.
So how do you get your catalog onto that shortlist of three? Start here.
As tempting as it is, you can’t be the best at everything … and AI won’t believe you if you try.
Pick the one or two things your product does better than anyone else. Then check that the proof is embedded nicely in your product data: how long it lasts, what it's made of, where it comes from, and which certifications it holds.
If the weight on your site doesn’t match the weight on a retailer’s listing, you’ve got a problem.
Keep copy, specs, images, and videos in one central place and push them out to every channel from there (ideally not from a spreadsheet).
AI reads your reviews as closely as it reads your product page. If there’s a contradiction between the two, it will prioritize what your customers have said. So, if you call a jacket waterproof and customers keep saying they got soaked, AI will believe the customers.
Keep an eye on what people are saying. When it doesn't match what you've written, update the copy or fix the product.
Type your shoppers' questions into ChatGPT or Gemini and see what comes back. Are you on the list? If not, who is, and where is the AI getting its information? Check back every few months, because the answers can and will change. Get your product data right everywhere, and you give both the AI and the shopper a reason to pick you.