Offloading Thoughts to the Bots: How Generative AI Access Influences Cognitive Reflection

Generative Artificial Intelligence (AI) systems are increasingly used to support reasoning and decision-making, raising questions about how AI influences reflective cognition and trust. This study examined how AI pedigree and reliability influenced cognitive reflection performance, cognitive load, trust, and performance overestimation during a verbal Cognitive Reflection Test (CRT-V). Participants ( N = 148) were assigned to a no-AI control condition, a low-pedigree AI condition, or a high-pedigree AI condition. Participants in the high-pedigree AI condition demonstrated significantly greater trust and CRT performance than participants in both other conditions while also exhibiting lower performance overestimation. AI access did not significantly reduce perceived cognitive load, although greater cognitive load predicted poorer CRT performance overall. Importantly, trust interacted with AI condition such that greater trust predicted lower CRT performance in the low-pedigree AI condition but not in the high-pedigree AI condition. These findings support theories of calibrated reliance rather than indiscriminate cognitive offloading.