How AI decides which products consumers see

For most of the internet era, online shopping has started with a product in mind. A shopper has an idea of what she wants, types in keywords, opens a dozen tabs, and compares specs and prices while slogging toward a decision. AI is now changing that, not only speeding up the research, but also creating a new consumer behavior. That’s because we don’t naturally think in terms of products. We think in pain points, goals, and constraints: “I need to look alive for my 9 a.m. meeting after my red-eye flight.” “I need ideas to get my toddler to eat veggies.” With GenAI tools like ChatGPT, Claude, and Gemini, shoppers no longer have to translate these needs into product searches. The toddler’s mom types in her query, and AI asks follow-up questions about the child’s favorite foods, texture preferences, and dietary restrictions to build a solution that includes relevant product recommendations. That compression of research into conversation moves commerce from a system that captures demand based on product searches to one that creates demand by recommending products shoppers might not even realize they need. With nearly half (45%) of consumers globally now shopping with AI, the discovery journey is moving upstream. But most brands haven’t followed it there and need to quickly ensure they meet the prerequisites to show up in AI recommendations. 2 STAGES TO RECOMMENDATIONS Many assume AI models simply take keywords from a user’s prompt and then generate a list of products whose descriptions refer to those keywords. But AI shopping isn’t keyword ranking; it’s a recommendation system built on semantics, constraints, and authority. To be recommended, a product actually requires clearing two hurdles, and stumbling over the first makes the second irrelevant. Stage 1 is entering the consideration set. Before any ranking happens, an AI model determines which products belong in the conversation. If a shopper asks for the best shampoo for a sensitive scalp under a certain price, the m