Cognitive Effects of ChatGPT: Is ChatGPT Making Us Forget How to Think?

Table of Contents

Your Brain on ChatGPT: What AI Writing Tools Are Doing to Memory, Creativity, and Critical Thinking

The rise of large language models (LLMs) like ChatGPT has revolutionized how we read, write, and learn. They’re fast, responsive, and seemingly intelligent—capable of generating long-form content with just a single prompt. But have you thought about cognitive effects of ChatGPT on our brain. A new 200-page research paper from MIT raises this serious concern:

What if AI writing tools are quietly eroding our ability to think, remember, and understand?

This blog dives deep into that study—titled Your Brain on ChatGPT: Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing—and explores the surprising cognitive trade-offs of relying on AI to write for us.

The Experiment: Three Types of Writers

To understand how LLMs affect our minds, the researchers divided participants into three groups:

Brain-Only Group – No AI, no internet, no tools. Essays were written from memory and personal knowledge.
Search Engine Group – Participants used traditional web search (like Google) to gather information.
LLM Group – Used OpenAI’s GPT-4.0 as their only information source.

Later, participants switched roles: AI users wrote essays without tools, and brain-only users tried using ChatGPT.

The researchers then analyzed brain activity, recall accuracy, writing quality, and participants’ sense of ownership over their work.

LLMs Make Writing Easier—But Shallow

One of the clearest findings was that using ChatGPT significantly reduces cognitive load—by 32% compared to traditional methods. It also increased productivity by 60%. AI users were more willing to engage with tasks and spent longer periods writing.

But this comes at a cost.

That lower mental effort also meant weaker memory formation, shallower understanding, and less critical thinking. Users often skimmed the surface of topics rather than building deeper cognitive maps or mental models.

In short, the brain wasn’t working as hard—and that’s not always a good thing.

Productivity vs. Comprehension

The study revealed a consistent trade-off:

LLM users produced more polished essays—grammatically perfect, well-structured, and coherent.
But they performed worse on follow-up memory tests and showed lower quality reasoning, especially on complex or scientific topics.

Why? Because LLMs streamline the information retrieval process. They offer neat, summarized responses that eliminate the need to sift through multiple sources. But that also means users do less active integration of information, which is essential for real understanding.

The Creativity Deficit

Another major discovery was around creativity.

Brain-only participants showed strong variation in how they approached essay writing. Their essays were diverse and unique.
LLM users, however, produced statistically homogeneous essays—similar in tone, structure, and even phrasing.

This points to a hidden danger: as more people use AI to write, our outputs may begin to converge, reducing originality and independent thought.

Can You Quote Yourself?

Participants were asked if they could recall a line from their own essay.

LLM users struggled—83% couldn’t remember a single accurate quote.
Search engine and brain-only participants had near-perfect recall by the third session.

This suggests that LLM-generated content simply doesn’t “stick.” Users don’t encode the information deeply because the mental work of writing and processing is being outsourced to the machine.

What About Brain Activity?

Using EEG scans, the researchers found:

Brain-only group had the highest levels of neural connectivity—especially in memory and planning regions.
Search engine users showed reduced connectivity (by 34–48%).
LLM users had the lowest—55% less brain activity during the writing task.

This isn’t just about memory or language—it’s about the very way our brains engage with tasks. Writing without help forced participants to be more creative, more analytical, and more cognitively involved.

Top-Down vs. Bottom-Up Thinking

The study identified two distinct mental approaches:

Bottom-up processing: Used by brain-only writers. They built their essays from smaller ideas, synthesizing knowledge through active exploration.
Top-down processing: Used by LLM users. They began with a complete or nearly complete idea from the AI and worked downward, refining and formatting.

This change in thinking style highlights a broader shift in how AI affects cognition: from creation to curation, from exploration to supervision.

The Ownership Problem

Another subtle but powerful insight: People don’t feel like they “own” the writing produced with AI help.

Brain-only and search groups reported full ownership.
LLM users gave mixed responses—some said they didn’t feel connected to their work at all.

This sense of disconnection may have long-term implications for how people feel about their learning, creativity, and autonomy.

Echo Chambers, Even in AI

We often hear about echo chambers on social media, where algorithms feed us more of what we already believe. But this paper points out that LLMs, too, can reinforce echo chambers.

Because they learn from dominant narratives and data patterns, they often return similar ideas, phrasings, and perspectives—especially if prompted in familiar ways.

This can subtly limit intellectual diversity and reinforce biases unless users make a conscious effort to push back.

The Scariest Finding: Lingering Cognitive Effects

Perhaps the most alarming discovery: the cognitive effects of AI linger even after you stop using it.

Those who started with ChatGPT and later wrote without it still underperformed on memory and comprehension.
In contrast, those who began by writing without tools and then used LLMs showed better learning outcomes overall.

In other words, early over-reliance on AI may impair your ability to process information deeply—even after the AI is gone.

Final Takeaways: Rethinking How We Use AI to Learn

This study challenges the idea that more productivity is always better. Here are the core lessons:

Use AI strategically, not as a shortcut. It’s a powerful tool, but not a replacement for deep thinking.
Start with your own brain. Use LLMs to expand, clarify, or refine—but not to generate ideas from scratch.
Don’t skip the struggle. Mental effort is how we build memory, understanding, and creativity.
Balance automation with exploration. Let AI help you—but don’t let it think for you.
Guard your individuality. AI can write for you, but it can’t replicate your personal insights, experiences, or voice.

The Bigger Picture

AI isn’t necessarily making us “dumber.” But it is changing the nature of how we think. Our brains are shifting from generators of knowledge to supervisors of machine-generated content.

If we’re not careful, we risk losing the very skills that make human intelligence unique: curiosity, creativity, and critical thinking.

The future will be defined not by who has access to AI—but by how wisely they choose to use it.

To read more about how AI is transforming learning check out our blog on AI’s impact on education.