AI research has moved from novelty to habit in the shopping journey, and it's changing more than where consumers start looking. It's changing who wins.

Some 63.1% of US consumers have used AI for product research, up from 49% a year earlier, according to a February 2026 Envision Horizons survey of 1,035 US adults. A third now use it weekly or more.

Consumers still start most searches on familiar ground, Google (49.2%) and Amazon (34.3%), but AI is embedded across both, through Google AI Overviews, Amazon Rufus (now named "Alexa for Shopping"), and a growing field of standalone assistants. What happens in their purchasing journey next has changed.

AI recommends. And vetoes.

Among consumers who shop with AI, 50.9% have abandoned a purchase after an AI assistant raised concerns, and 12.1% said it happens often. The same tool that recommends products also eliminates them: 40.7% of AI shoppers say its suggestions are "often" or "always" well-suited to their needs. That dual role, endorser and gatekeeper, is what gives AI's veto power teeth.

The veto isn't arbitrary. It tracks to what consumers ask AI to check, according to the report.

Some 37.7% use it to summarize reviews, and AI surfaces the pattern behind the star rating, not just the number, so a product with 4.2 stars and recurring complaints about durability gets flagged anyway. Another 22.1% ask AI to parse technical specs, and vague or inconsistent spec sheets give the model less to work with, pushing it toward competitors whose data is cleaner. A smaller share, 18.3%, ask about ingredient safety and effectiveness, a category where unsubstantiated claims either go unverified or actively raise doubt.

The thread connecting all three: AI doesn't reject products on judgment. It rejects them when the data can't be verified.

Trust varies sharply by category, and brands need to read the gap.

Consumers will hand AI more control in some categories than others. Some 36.7% are comfortable buying everyday household items primarily on an AI recommendation, and electronics follow at 33%, according to Envision Horizons' survey. Pet products drop to 17.1%. Baby products bottom out at 5.8%, a sixfold gap between the most and least trusted categories.

That gap sets the bar differently by category. In low-risk purchases, thin product data gets a brand vetoed fast and replaced by whoever documented better. In categories carrying emotional or safety weight, strong data isn't enough on its own. Consumers want reassurance from multiple sources, reviews, community, and the brand's own site, before they'll let AI settle it. Notably, 38.1% of consumers said they always do their own research regardless, a reminder that AI is an input for most shoppers, not a verdict.

Every AI query is a chance to lose the sale to someone else.

Some 30.1% of AI shoppers use the tool specifically to find alternatives to something they're already considering, effectively inviting a competitor into the room before they've decided anything. That habit shows up in outcomes: 79% of AI shoppers have switched brands because of an AI recommendation, with 35.5% saying it happens sometimes and 14.6% often or almost always.

Brand name isn't holding the line. When AI presents options side by side, price drives 34.7% of decisions, star ratings drive 26.2%, and familiarity drives 12.2%. Brand name alone accounts for just 9.8%. Every AI recommendation is a forced comparison, and the product with better documentation tends to win it regardless of how well known the brand is.

Where consumers are having those comparisons is shifting fast. ChatGPT still leads AI shopping usage, growing from 35% to 40.9% YoY, but Google Gemini posted the largest jump, from 26% to 38.5%, closing in on ChatGPT's lead. Amazon Rufus grew 50%, from 9% to 13.5%, and Walmart Sparky, new to the survey this year, debuted at 7.5%. Consumers aren't consolidating around one assistant. They're spreading across several, which means brands need visibility on more than one platform to stay in the conversation.

Amazon is winning the checkout even when AI wins the research.

Among AI shoppers, 50.9% completed their most recent AI-researched purchase on Amazon, according to the study. Retailer sites like Target and Walmart captured 17.6%, in-store purchases took 15.6%, and brand D2C sites closed out just 7.8%, meaning Amazon converts nearly seven times more AI-influenced purchases than brand-owned channels do. Amazon's conversion advantage on purchase-intent traffic, an estimated 12 to 15% versus 1 to 2% on D2C sites, isn't something AI research disrupts. It reinforces it. A consumer can discover a product through AI, validate it on the brand's own site, and still check out on Amazon.

D2C traffic hasn't disappeared, though. Some 14.1% of AI shoppers click through to a brand's own site after AI research, not to buy but to confirm what AI told them, check claims, or browse the full product line before converting elsewhere. That pattern shows up most in categories like baby and pet, where consumers want the brand's own voice as a final check before they trust AI's read. D2C's job in an AI-mediated funnel isn't closing the sale anymore. It's being the source of truth AI can point to and consumers can verify against.

This was originally featured in the Retail Daily newsletter. For more retail insights, statistics, and trends, subscribe here.