People are using AI assistance more often than they realize, and they overestimate how much time it actually saves them.

What happened: Researchers conducted three studies with 2,691 participants and found that people frequently choose to use AI for simple tasks like arithmetic and spell-checking even when doing so provides no meaningful time or effort savings. Participants also significantly underestimated how much they were actually using AI.

Why it matters: The studies reveal two systematic biases in how people think about AI use. First, people believe they use AI less than they actually do. Second, they overestimate the time and effort savings AI provides. These miscalibrations create what the researchers call an "overreliance feedback loop"—where using AI once leads to more AI adoption, which then entrenches incorrect beliefs about the time savings involved.

What to watch: The research identifies a session-level carryover effect, meaning that a participant's prior AI use directly leads to further AI adoption. This suggests that even modest AI use can shift people's behavior in ways they do not accurately perceive, raising questions about whether people's reliance on AI tools is actually benefiting their productivity as much as they think.