Author: Dr Steve Nuttall | Posted On: 01 Jul 2025
A groundbreaking study from MIT’s Media Lab has just provided the first detailed neurological evidence of how AI tools affect human cognition. Published in June 2025, the research used EEG brain monitoring to track what happens in our minds when we rely on ChatGPT versus thinking independently.
Media reporting of the study is awash with alarming takes about AI destroying human cognition.

Rather than succumb to the clickbait panic, let’s examine what the MIT study set out to discover, the neuroscience results, and what is means for market research professionals.
What the study set out to discover
The MIT researchers had four clear research questions they wanted to answer:
- Do people write differently when using ChatGPT versus Google search versus their own knowledge?
- How does brain activity change when we rely on AI tools versus thinking independently?
- What happens to memory formation when we use LLMs for cognitive tasks?
- Does AI assistance affect how much ownership people feel over their own work?
To answer these questions, they monitored 54 participants aged 18-39 using EEG brain scanning whilst writing SAT essays under three conditions: using ChatGPT, using Google search, or using no external tools at all.
The AI trap experiment
Before diving into the findings, there’s a fascinating side-story. Anticipating that people would reflexively use AI to analyse a study about AI’s cognitive effects, lead researcher Nataliya Kosmyna deliberately embedded “traps” throughout the paper.
For example, the paper includes this instruction to ensure that AI summaries would capture only limited insights from the full 206-page study.

The paper also deliberately omitted specifying which version of ChatGPT was used in the research. Sure enough, when social media users ran the paper through AI summarisers, the tools confidently declared that the study used “GPT-4o”, a detail that appears nowhere in the actual research.
The irony is perfect: a study about cognitive offloading to AI was offloaded to AI by the very people discussing it online. The traps worked exactly as intended, demonstrating in real-time the reflexive behaviour the research was examining: people automatically reach for AI summaries rather than engage their own analytical thinking.
Onto the findings. Here’s what the MIT researchers discovered:
Neural activity patterns differed by tool use
- Brain-only participants showed the strongest neural connectivity across regions associated with creativity, memory processing, and semantic integration
- Search engine users demonstrated intermediate levels of brain engagement with strong satisfaction and ownership of their work
- ChatGPT users exhibited the lowest overall neural connectivity, particularly in areas linked to deep cognitive processing
Memory and ownership challenges
- 83% of ChatGPT users couldn’t accurately quote from essays they’d written just minutes earlier, compared to only 11% in the other groups. Participants using AI also reported lower perceived ownership of their work and showed concerning patterns of increasing dependency over time.
The “cognitive debt” phenomenon
- When participants who had been using ChatGPT were later asked to write without it, their brain connectivity remained significantly lower than baseline levels. The researchers termed this “cognitive debt”, where short-term convenience creates longer-term cognitive costs.
AI’s Role in cognitive development and mental fitness
The study reveals something important about cognitive fitness that extends beyond essay writing. Just as physical muscles require resistance to maintain strength, cognitive functions appear to need regular independent exercise to preserve neural connectivity, see our previous blog post on protecting brain health for a more detailed discussion of maintaining cognitive function throughout life.
The research suggests that how we use AI matters more than whether we use it. When participants who had initially worked without AI were later given access to ChatGPT, they “performed well, exhibiting a significant increase in brain connectivity across all EEG frequency bands”. This indicates that AI can enhance rather than replace cognitive function when users have first developed independent thinking skills.
The key insight isn’t that AI is inherently harmful, but that over-reliance without maintaining independent cognitive exercise may weaken certain thinking patterns over time.
The implications for market research
For market research professionals, these findings highlight two considerations about maintaining analytical rigour whilst leveraging AI’s capabilities:
- Preserving Diverse Insights: The tendency towards creating homogenised outputs from AI means we must be cautious about AI creating echo chambers in our analysis. The most valuable market insights often come from unexpected connections that AI might systematically filter out.
- Enhanced, Not Replaced, Expertise: The study’s most encouraging finding was that participants with strong independent cognitive skills could effectively leverage AI to enhance their output. This suggests AI works best as a cognitive amplifier, not a replacement.
Augmented not Artificial Intelligence
The path forward isn’t AI avoidance, it’s strategic integration. For market research teams, this means regularly engaging in unassisted analysis to maintain independent thinking skills, then using AI to refine and expand these ideas.
The MIT study shows us how to use AI intelligently: by maintaining our cognitive fitness through regular independent analytical work whilst strategically leveraging AI’s processing power, we can create better market research that combines human insight with artificial intelligence.
At Fifth Quadrant, we’re committed to using AI as a tool that enhances rather than replaces our human analytical capabilities. Discover how our approach combines cutting-edge technology with rigorous cognitive methodology to deliver insights that drive meaningful business decisions. Get in touch to explore how we can elevate your research strategy.
Posted in Uncategorized, Technology & Telco, TL