The Maze: AI shopping tools can help people discover a product, compare alternatives or assemble a basket. Those jobs require different levels of trust. A December 2025 survey covering France, Germany and the UK puts comparison ahead of ordering household staples again. For brands, that raises a practical question: can an assistant explain why their product belongs on the shortlist before a shopper reaches their website?
Comparison is the strongest task signal. McKinsey’s survey reports 63% for comparing options, 55% for learning about products and 46% for discovery or inspiration. These are responses about AI use during the previous three months. They describe several possible shopping tasks, rather than slices of one market. Comparison can include checking models, prices, quality and reviews. A brand therefore needs more than an attractive product name: an assistant must be able to identify the relevant specifications and explain the trade-offs. That is an implication of the task pattern, not proof that better data will automatically increase sales.
Basket assistance and repeat orders are different propositions. Building or optimizing a basket records 33%; repurchase or replenishment records 21%. Basket assistance can mean finding compatible items or a cheaper substitute. Replenishment means ordering familiar products again. Neither figure measures completed autonomous payments. Comparison exceeds replenishment by 42 percentage points, but this is no checkout abandonment rate: the responses do not follow one group through successive steps. Treating them as a sales funnel would turn a useful distinction between tasks into an invented loss of customers.
Support breaks the story of a steady decline. Postpurchase support records 39%, above basket work at 33%. People may need help resolving an order problem, starting a return or understanding a product after delivery. AI assistance can therefore matter on both sides of payment without taking control of payment itself. For a retailer, accurate instructions and accessible order policies could be useful alongside product descriptions. The evidence supports testing those jobs separately. It does not show that a single assistant should handle everything or that shoppers have accepted a standing mandate to buy for them.
The denominator deserves as much care as the headline. The research firm surveyed 749 people: 102 in France, 401 in Germany and 246 in the UK. Its task exhibit prints that sample size, yet its overall measure puts AI use for shopping research or purchase decisions at 38%. The accessible report does not clearly explain how the task questions were routed. The 63% comparison response should therefore remain a task figure, rather than a claim that 63% of all shoppers use AI to shop. These are self-reported answers from December 2025, not current transaction logs or evidence covering every European country.
Why it matters: A retailer can make its products easier to compare while leaving the purchase decision with the customer. Clear specifications, credible reviews and accurate availability give an assistant something useful to evaluate. The same operational lesson is worth testing elsewhere, but the survey does not establish adoption outside these three markets. Measure whether shoppers get better answers and find suitable products before promising a robot sales force.
