The Criteo 2026 Commerce & AI Trend Report, conducted across January and February 2026, explored digital and omnichannel shopping habits of consumers across various markets and categories. AI is already having a major impact on discovery, with 47% of consumers surveyed using it for comparison, 39% for finding new products and 38% for finding deals – but only 8% use AI to actually place an order, showing AI’s role at the checkout stage is still limited.
AI-assisted shopping can create an echo-chamber, but customers will always want exciting discovery experiences – 56% of shoppers would like product recommendations beyond exact search matches – so the future lies in tailored recommendations combined with relevant surprises.
96% of shoppers who use AI shopping assistants regularly also use other channels along the way (such as search, social, brand websites or retailer websites), and only 14% start their shopping journey with an AI assistant. 36% said marketplace was their first go-to, 28% search engine, and 15% retailer website.
The survey also found that while privacy is a consideration during AI-assisted shopping – when it comes to sharing information, 52% said it depends on the topic and 25% are generally comfortable if it improves the experience. Consumers are most cautious about sharing personal identity or location data (57%) and payment-related information (55%).
It is also clear just how important trust is – with the results showing it is a main requirement for AI-assisted shopping. 52% of consumers said they want to be protected from fake or biased content.
Andy Stephen, managing director, retail media UK at Criteo, said: “AI assistants that can surprise and delight, delivering options that preserve the sense of exploration in online shopping, will become the gold standard. Far from being random, these ‘relevant surprises’ should be based on robust recommendation algorithms. As a result, brands will need to start marketing to agents as well as people. Success won’t only hinge on visibility in feeds or search results, but also on how clearly a brand is understood by AI systems – when it’s the right answer, how it fits into bundles, and whether it’s trusted.”
