# AI in education risks critical thinking, study finds

**Published:** 2026-08-06T16:01:35.213Z  
**Topic:** Google Ai  
**Sentiment:** neutral  
**Publisher:** TrendWatcher — https://www.trendwatcher.in/article/287f4051-fe26-4131-9789-9beff2530c1d

AI education study shows 20% of ChatGPT chats are learning‑related, 40% of users under 24, and two‑thirds of students say AI harms critical thinking.

AI‑driven tools are already reshaping how students learn: a RAND survey of 1,200 students reports that two‑thirds believe AI is eroding their critical‑thinking abilities, while OpenAI data show roughly 20% of ChatGPT conversations involve education topics and about 40% of its users are under 24 years old [2].  

| At a glance | |
|---|---|
| Education‑related chats | ~20% of ChatGPT conversations |
| Users under 24 | ~40% of ChatGPT users |
| Student perception | 66% say AI harms critical thinking |
| RAND survey size | 1,200 students |

## AI output vs. learning process  

The Conversation warns that large language models (LLMs) produce fluent, confident answers without a built‑in commitment to truth, allowing learners to bypass the effort required for genuine understanding [1]. An MIT Media Lab study cited there found that students who wrote essays with LLM assistance showed the weakest brain connectivity and could not recall their own text, indicating that the cognitive work of learning was outsourced. This “epistemic atrophy” – the gradual loss of the habit of questioning and verifying – mirrors historic concerns about “epistemic injustice,” but now manifests as a decline in the willingness to engage with uncertainty.

## Institutional responses and limits  

OpenAI’s vice‑president of education, Leah Belsky, distinguishes between AI that “avoids thought” and AI that “extends thought,” arguing that the former removes the productive friction essential for durable learning [2]. OpenAI’s Study Mode, launched last year, attempts to preserve that friction by prompting follow‑up questions rather than delivering instant answers. However, Belsky notes that any product setting can be sidestepped; students can simply start a new chat outside the guided environment, underscoring the need for curricula that teach “productive difficulty” and the ability to recognize when convenience becomes evasion [2].

## What to watch  

- **OpenAI Study Mode adoption** – monitor usage metrics and any updates that tighten the follow‑up questioning loop.  
- **University pilots** – track results from Harvard’s “flipped case” and Duke’s AI‑enhanced group‑work analyses for evidence of scalable impact.  
- **Policy discussions** – watch for education‑sector guidelines that address AI’s role in assessment and the preservation of critical‑thinking skills.  

The emerging evidence suggests that while AI can accelerate information access, it also threatens the deep cognitive habits that underpin true learning. Whether educational institutions can redesign curricula to retain the “productive difficulty” essential for thinking remains an open question.

## Sources
1. The Conversation — [AI doesn’t just help us think, it thinks instead of us: what this means for the process of learning](https://theconversation.com/ai-doesnt-just-help-us-think-it-thinks-instead-of-us-what-this-means-for-the-process-of-learning-284038)
2. Forbes — [AI In Education: Forming Citizens When AI Can Do The Work](https://www.forbes.com/sites/rayravaglia/2026/07/23/ai-in-education-forming-citizens-when-ai-can-do-the-work/)

---
Cite as: TrendWatcher, "AI in education risks critical thinking, study finds", https://www.trendwatcher.in/article/287f4051-fe26-4131-9789-9beff2530c1d