Origin
"AI overreliance" is the term cognitive science and education research uses for what colloquial coverage calls AI dependency. The term is more measured because it doesn't borrow from clinical vocabulary it doesn't belong to. The underlying pattern is the same: AI use past the point of useful assistance, where the tool starts substituting for the cognitive work rather than supporting it.
The strongest single research signal is the 2024 Springer paper on AI dialogue systems and student cognitive abilities, which measured what happens when AI assistance is the default learning surface. The findings were consistent with the 2025 Gerlich study covered by PsyPost: heavier AI use correlated with weaker critical-thinking measures, mediated by cognitive offloading.
The split that matters
EDUCAUSE Review's late-2025 framing of the "paradox of AI assistance" captures the key distinction cleanly: AI is net-positive when used as a thinking partner (conceptual inquiry, brainstorming, structured questioning) and net-negative when used as a thinking substitute (drafting in full, generating reasoning chains, producing finished output without engagement).
The same split shows up in the Anthropic 2026 developer study: developers who used AI for conceptual inquiry scored 65%+ on comprehension tests; those who used it for code generation scored under 40%. Same tool, different uses, very different outcomes.
What overreliance looks like in different domains
The pattern is consistent across cognitive domains:
- In writing, overreliance is letting AI draft the first version of pieces you'd be asked to defend. The encoding gap is largest in first-draft territory.
- In coding, overreliance is shipping AI-generated code without reading it line by line. The VirtusLab "cognitive debt" framing captures this well.
- In reading, overreliance is reading the AI summary of a piece you would benefit from reading yourself. The summary is information-shaped but the encoding isn't happening.
- In reasoning, overreliance is taking the AI's reasoning chain as the answer rather than as input to your own reasoning.
How to recalibrate
The four habits that show up across the strongest published advice:
- AI for conceptual inquiry, not just generation.
- Read everything AI gives you before shipping it.
- Set AI windows so the tool isn't always on.
- Keep daily unmediated practice somewhere on the calendar.
These are the same four moves recommended for AI dependency and for AI brain fry — the underlying mechanism is the same in all three framings, just at different layers of the experience.
In Senwitt
Senwitt's role in the AI-overreliance picture is the fourth habit: the daily unmediated practice. The daily Set is short, mixed across six Skills, and entirely off the AI surface. It is not a treatment for overreliance — overreliance is a workflow pattern, not a clinical condition — but it is the daily-practice piece of the published recalibration advice.
Related concepts
- AI dependency — the colloquial label for the same pattern
- AI brain fry — the cognitive-fatigue cousin
- Cognitive offloading — the academic frame
- Cognitive debt — the MIT-era variant
- Skill atrophy — the long-term skill consequence
- AI overreliance research page — full research treatment
