Why AI reliance may undermine confidence
Mastery experiences are the felt successes that build self-efficacy—the belief that you can handle a task. When AI does the cognitive work for you, you may skip the productive struggle that creates those successes. One study proposes that AI dependence may reduce mastery experiences and self-efficacy, and that this pathway helps explain how academic stress leads to burnout and anxiety [1]. The same study notes that initial self-efficacy might also drive greater AI dependence, suggesting the relationship could run in both directions [1].
Among medical students, AI dependency has been examined alongside self-efficacy and academic stress, though the abstract does not specify the direction or strength of the association [3]. In tertiary students, perceived AI ability appears to influence academic self-efficacy through independent learning disposition, meaning how students approach learning on their own may be the bridge between AI use and confidence [4].
Does academic stress make the effect stronger?
The clearest evidence for stress as an amplifier comes from the study that places AI dependence in the middle of a pathway from academic stress to burnout and anxiety [1]. That framing implies that when stress is high, people may lean more on AI, and that leaning may then erode the mastery experiences that would otherwise buffer stress. However, the abstract does not report a direct test of stress as a moderator, so the stronger-when-stressed claim is an inference from the mediation model rather than a confirmed interaction [1].
Other studies treat academic stress as a covariate to control rather than a factor that changes the AI–self-efficacy relationship [4]. One study of nursing students links AI-related constructs to self-efficacy but does not examine stress as a moderator [5]. So the answer to whether stress strengthens the effect is a qualified yes: the pathway is consistent with that idea, but no study directly compares high-stress and low-stress groups on this relationship.
What this means for using AI well
If AI replaces the struggle that builds mastery, confidence may suffer—especially under pressure. But the evidence also hints that AI can support self-regulated learning, which involves planning, monitoring, and adjusting your own learning [5]. The difference may lie in whether AI is used as a tool that scaffolds your thinking or as a substitute that does the thinking for you.
For a student or teacher, the practical takeaway is to watch for signs of dependence: using AI before attempting a problem, feeling unable to start without it, or losing track of how an answer was reached. These are the moments when mastery experiences may be skipped. The research does not yet show a causal chain from AI use to lost self-efficacy, but the proposed pathway [1] and the associations found in other studies [3][4][5] suggest that keeping some productive struggle in the learning process is worth protecting.
Sources used in this answer
When cognitive offloading becomes dependence: how AI dependence mediates the pathway from academic stress to burnout and anxiety
AI dependence may mediate a pathway from academic stress to burnout and anxiety by reducing mastery experiences and self-efficacy, though initial self-efficacy might also drive greater AI dependence.
A Three-Wave Longitudinal Study on Academic Self-Efficacy, Technostress, and Learning Outcomes among French Educators
The study examines academic self-efficacy, technostress, and learning outcomes among French educators using generative AI, with a focus on mastery experiences, but does not clearly address academic stress as a moderator.
Association of self-efficacy, stress, performance expectation, and procrastination with AI dependency among medical students at a public sector university in …
The study examines associations between AI dependency and self-efficacy among medical students while considering academic stress, performance expectation, and procrastination, but the abstract does not specify the direction or strength of these associations.
Exploring the relationships among perceived AI ability, academic self-efficacy and independent learning disposition in the tertiary contexts
Perceived AI ability appears to influence academic self-efficacy through independent learning disposition in tertiary students, with academic stress treated as a covariate rather than a moderator.
AIlessphobia, AI-supported self-regulated learning, and academic self-efficacy in nursing students: A structural equation modeling approach
AI-related constructs, including AIlessphobia and AI-supported self-regulated learning, are linked to academic self-efficacy in nursing students, but academic stress is not explored as a moderator.