Cognitive Offloading in the Age of AI: How Bot Typing Speed and Math Skill Shape the Choice to Offload

Generative artificial intelligence (AI) is increasingly integrated into everyday problem solving, yet still too little is known about how system characteristics and user abilities shape cognitive offloading to such tools. In this preregistered experiment, we investigated whether the responsiveness (i.e., the typing speed) of an AI simulated by a bot and participants’ math skill predicted how much they relied on the AI when solving arithmetic problems. A total of 198 participants completed a choice/no-choice paradigm in three blocks (“forced internal”, “forced external” and “choice”). Results revealed that participants in the “fast bot”-condition relied more on the AI than those in the “slow bot”-condition. Further, higher math skill predicted less offloading. The interaction between bot speed and math skill was not significant. Exploratory analyses with adjusted exclusion criteria and clipped data replicated these results. Together, the findings extend research on cognitive offloading to the context of generative AI, suggesting that even minor interface parameters, such as typing speed, can alter the balance between internal and external strategies. These results highlight the importance of considering both tool design and user characteristics in promoting effective and sustainable AI use, particularly in educational contexts.