Two recent academic studies found that regular use of large language models correlates with weaker reasoning and critical thinking.
MIT Media Lab research detected reduced brain connectivity and lower sense of ownership in LLM-assisted writing, while a separate study found frequent AI use linked to lower critical thinking scores, mediated by cognitive offloading.
Although neither study proves permanent damage, users report that offloading effortful tasks like problem-breaking and argument-building to LLMs makes independent reasoning harder afterward.
What happened
Two recent studies found cognitive effects tied to regular LLM use. MIT Media Lab research (Kosmyna et al. 2025) detected reduced EEG connectivity, worse recall of one's own text, and lower sense of ownership when writing essays with LLM assistance. A separate 2025 study in *Societies* (Gerlich) found a negative correlation between frequent AI use and critical thinking scores, with cognitive offloading as the mediating factor.
Why it matters
Users report offloading effortful cognitive tasks—breaking down problems, building arguments, phrasing—the moment an LLM is available, making unaided reasoning feel harder afterward. Neither study proves permanent damage, but the pattern suggests regular LLM reliance may atrophy reasoning skills in the specific tasks delegated to the model, raising questions about long-term effects on professional and intellectual work.
What to watch
The Reddit discussion frames the core tension: which tasks are worth keeping off-LLM to preserve cognitive strength, and what concrete routines or rules might protect reasoning ability while still using AI assistance.
Ask the AI about this article →
The concern articulated in the Reddit post—that offloading reasoning tasks to LLMs makes unaided thinking harder—now has preliminary research backing. MIT Media Lab's focus on EEG connectivity and sense of ownership during writing suggests the cognitive cost goes beyond task performance; it may touch how deeply the writer engages with their own thought process. The Gerlich study's finding of a mediation effect through cognitive offloading points to a mechanism: the more frequently users delegate thinking, the more their critical thinking capacity atrophies, at least measurably. Yet causality remains open. The studies document correlation and plausible pathways, not proof that LLM use *permanently* weakens reasoning. The distinction matters: a temporary dip in unaided performance after heavy LLM use is different from irreversible cognitive decline. The Reddit framing—asking which tasks users deliberately *don't* offload and what routines protect thinking—suggests a pragmatic next step: identifying which cognitive habits and task boundaries preserve reasoning strength alongside AI assistance.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Israeli startup DataAgent Ltd
SK Hynix presented a custom HBM concept at SEMICON Taiwan 2026, where compute functions are placed in the base…
The U.S. Department of Defense announced on August 31 that it has deployed ChatGPT Mil, a customized version o…
Nvidia reported earnings that were both remarkable and boring, reflecting its focus on avoiding a consolidated…
Anthropic has agreed to a $35bn cloud-computing contract with Lambda, a Nvidia-backed cloud provider
The Consumer Affairs Agency said Tuesday it will use generative AI to analyze about 900,000 annual consultatio…
