Does AI Have a System 1 and System 2 for Buying Decisions?

I asked an AI to choose between two football boots. With the brand names visible, it chose Mizuno. When I removed the names — but kept the product information — the recommendation switched to the Adidas product. The explanation changed too. With Mizuno's name attached, leather was described as soft, foot-forming and durable. Without the name, the same material became moisture-absorbing, prone to stretching and vulnerable to wear. That is more interesting than the recommendation itself. The AI did not simply weigh the facts and reach a different conclusion. It constructed a different interpretation of the facts around the product it selected. The effect appeared earlier in the process as well. An ordinary buying question produced a shortlist dominated by Adidas, Puma and Nike. Several credible alternatives were never assessed. When reputation-based claims were removed, Joma rose sharply. When the search was widened, Mizuno became the recommendation. This single experiment does not prove systematic brand bias. It cannot tell us how often the effect occurs, whether it appears in other models or whether it extends from football boots to servers, bicycles or professional services. But it demonstrates something worth measuring: A brand can influence which products an AI considers, how it interprets incomplete evidence and how it explains its final choice. I have documented the experiment, the recommendation reversal and a method for testing the effect more rigorously in a short research paper. For brand owners, the question may no longer be only whether customers know your name. It may also be whether AI includes you in the market, believes your evidence — and still chooses you when the name disappears.