Generative AI Reliance and Cognitive Capacity in Professional Populations: A Descriptive Study of Interaction Style, Confidence, and Participant-Reported Autonomy

Generative artificial intelligence (GenAI) systems have rapidly become embedded in everyday cognitive workflows, raising questions about how reliance on these tools relates to executive engagement and perceived cognitive autonomy. This non-clinical mixed-methods study examines behavioral and self-reported patterns of cognitive offloading among high-use GenAI users.A total of 1,923 adults across Canada (n = 946) and the United States (n = 977), ages 25–57 (M = 38.4), completed ten GenAI-assisted tasks designed to simulate common executive demands, including planning under uncertainty, multi-step sequencing, decision-making, and reflective reasoning. Participants were encouraged to use commercially available large language model systems as they normally would. Measures included prompt activity, time-to-decision, override behavior, and Likert-scale assessments of AI reliance, confidence in reasoning, and perceived cognitive autonomy.Across tasks, 58% ± 7% of participants agreed that “AI did most of the thinking,” with higher reported offloading during planning and sequencing tasks. Greater prompt dependence and lower override frequency were associated with reduced self-reported confidence in independent reasoning (r = −.61, p < .01). Qualitative responses highlighted recurring themes of cognitive outsourcing, diminished perceived ownership of ideas, and trade-offs between speed and depth of thought.Findings are descriptive and do not support causal inference. They document variability in user engagement with GenAI during executive tasks and suggest that interaction patterns - such as active oversight versus unequivocal output acceptance - may be relevant to perceived cognitive authorship and autonomy, motivating future longitudinal and generative AI design-focused research.