If you feel like the world around you is getting dumber, it may not be your imagination. And if you rely on AI to help order your thoughts, you may no longer have an imagination to speak of anyway.

Since ChatGPT exploded into the public consciousness in late 2022, the push for AI adoption has been relentless. Who needs novelists when AI can write whatever your heart desires in a matter of minutes? Who needs friends when AI can listen patiently while you gaslight yourself into believing you are the centre of the universe? 

AI will solve the climate crisis if you feed it enough energy and water. It will solve cancer without paying a single doctor and it will reduce the court backlog without mending a single roof. AI hype promotes magical thinking while undermining our capacity for critical thinking in a very concrete way. 

What GPS tells us about cognitive offloading

One of the astounding things about humanity in the 21st century is our apparent inability to learn from experience. While AI in our pockets is relatively recent, the incorporation of GPS into our smartphones over the past couple of decades has already had an impact on our ability to find our own way in the world.

Research shows that habitual reliance on GPS leads to a progressive decline in users’ spatial memory

Researchers from UCL studying the brains of participants navigating the streets of Soho in 2017 observed that “the hippocampus simulates journeys on future possible paths while the prefrontal cortex helps us to plan which ones will get us to our destination. When we have technology telling us which way to go, however, these parts of the brain simply don’t respond to the street network. In that sense our brain has switched off its interest in the streets around us.” 

More recent studies indicate a worrying long-term impact. Research reported in Nature in 2020 showed that habitual reliance on GPS led to a progressive decline in users’ spatial memory. It is not simply the case that people with no sense of direction rely on GPS, but that reliance on GPS apparently causes us to become lost without it. Studies have shown that taxi drivers relying on their own knowledge of the streets and navigational skills have a larger posterior hippocampus than control groups. Our brains literally change shape depending on how we use them. And if you don’t use it, you lose it.

Our overuse of technology has effectively limited the way we look at the world around us, setting up roadblocks on the paths less travelled and reducing our ability to think for ourselves. If we looked up from our phones, we could see the warning signs. Social media has had a significant impact on our communities, our democracy, and our inner lives, yet we seem incapable of perceiving the risks riding the next wave of society-defining technology. Large language model (LLM) chatbots, for example, appear to understand us and know everything about our world. We’re blinded by the light of this shiny personalisation, but what impact is it having on our minds, and what will it mean for future generations?

What the ‘Your Brain on ChatGPT’ study found

In June of 2025, researchers from MIT published a revealing study called “Your brain on ChatGPT: accumulation of cognitive debt when using an AI assistant for essay-writing task”. They had looked at the neural and behavioural consequences of LLM-assisted essay writing on a group of participants that had been divided into three groups: LLM, search engine and brain-only. The groups undertook essay-writing tasks while their brains and cognitive load were assessed using electroencephalography (EEG). The essays were then scored using natural language processing, human teachers and an AI judge. You might be sceptical of an AI judge, but the results of this study should give pause for thought – while we are still capable of it. (Subsequent research has added more nuance to these findings; see the September 2026 at the end of the article.)

The EEG showed marked differences in brain connectivity between the groups, but strong similarities within each group. The strongest, most distributed neural networks showed up in the brain-only participants. Search-engine users showed moderate engagement, while those relying on LLMs had the lowest levels of all. It seems that when you use your brain, it lights up, but when you use AI, it remains in the dark. That might not be surprising – after all, if you’re using an LLM to write an essay, you’re not really using your brain at all. But the impact of using LLMs over time is not the same as giving your brain a bit of R&R: it has debilitating consequences.  

In a second stage of the experiment, researchers swapped the groups around, asking the LLM participants to conduct the task using their brain only and the brain-only participants to do it using an LLM. Those who had used an LLM in the first stage of the experiment showed reduced connectivity and could not engage effectively with the task. The brain-only group showed higher memory recall and brain activation when using the LLM, as did the search-engine group. Conclusion: reliance on AI seems to make our brains lazier, even when the AI is taken away.

The behavioural data produced by the experiment was equally revealing. When tested on their essays afterwards, none of the LLM cohort could quote their work correctly, though only 83% recognised the problem, which persisted over subsequent sessions. For the 17% who incorrectly believed they could quote their work, a bit of AI-fuelled imposter syndrome would not have gone amiss. By contrast, brain-only and search-engine participants were able to quote their work perfectly by the third post-essay session. Conclusion: not only does AI not understand what it produces in any real way, neither do its users.

AI, memory and ownership of our work

While attentive editing of a text is important for clarity and flow, the eureka moments when writing a piece of work usually happen in the bubbles of inspiration flowing from the ink in your pen, the rhythmic tap of your fingers on the keyboard or even from the steam rising out of the kettle as you pause to think. Writing is a process that is linked to our physical and mental reality. If you short-circuit it with AI, you eliminate that thought process and risk being left with nothing but empty words.

Those who write essays using AI feel almost no sense of ownership of their work, according to the research. If you don’t own it, you don’t understand it, and if you can’t remember it, it probably wasn’t worth doing in the first place. This is also true of those AI-prompted emails or for the Meta AI message you sent to the date you’ve effectively ghosted.

If you include expletives in your search terms, AI will apparently eff off and leave you to it

Of course, like any self-respecting researcher, I did ask ChatGPT to comment on the cognitive impact of AI on humans. And like any modern politician, it hedged its bets, accepting there might be adverse effects but stressing that AI can enhance critical thinking, and including a helpful bullet-point summary. I was too bored to read to the end.

AI advocates say it’s important to check, perfect and personalise any AI-generated text. But if you can’t be bothered to write it in the first place, are you really going to put the cognitive effort into that? It is also virtually impossible to factcheck anything without relying on utterly unreliable AI summaries in search engines. Thankfully, there is a hack. If you include expletives in your search terms, AI will apparently eff off and leave you to it. 

Perhaps this is proof of recent scientific findings that swearing is a sign of high intelligence. Apparently, it can help with pain management, emotional regulation and stress relief – all of which will no doubt come in useful for coping with the AI apocalypse when it comes. Swearing apparently boosts creativity and problem solving. Well, thank fuck for that! 

Of course, I’m not a scientist, this is not a peer-reviewed academic paper and I may be biased. You should do your own research, ideally without the help of an AI search engine. But if you are worried that refusing AI tools will render you obsolete, think again. When the rivers run dry, the data centres overheat and the dust settles on an AI-induced Armageddon, those who have refused to embrace the technology may be the only survivors still able to think their way out of it. 

So, if you want to future-proof your children, don’t teach them how to prompt AI but to resist, and then send them out on Duke of Edinburgh survival challenges in Soho without a phone. You probably don’t need to teach them to cuss, IG they’ve already got that…

Susie Alegre is a barrister and the author of “Freedom to Think” and “Human Rights, Robot Wrongs: Being Human in the Age of AI”