My buddy Claude, friend or foe?

July 21st, 2026

Written by Victoria Subritzky Katz

You might have used generative artificial intelligence (AI) today without thinking twice about it, asking it to draft an email, explain a concept, or relying on it to summarize your google search results instead of opening the links yourself — and you wouldn’t be alone. A billion people are monthly users of ChatGPT’s mobile app¹ and the rate of adoption for the new technology is out pacing the adoption of the internet², back when it was the big technological revelation. But while generative AI is doing the heavy lifting, what’s happening (or not happening) inside your skull?

The generative AI technology behind ChatGPT, Claude, and Gemini are large language models (LLMs), which are pattern-recognition systems trained on enormous amounts of text to predict, word by word, what a fluent response should look like. This means they don’t think or understand the way a brain does, but rather they generate humanlike answers by recognizing patterns learned from billions of sentences.

We know what’s going on inside the LLMs “mind” when we ask it to think through a complex problem, but do we shortchange our own minds by not thinking through that complex problem ourselves? 

The case for foe: the use it or lose it philosophy of the brain

Years of dancing ballet when I was younger strengthened my calf muscles, but I’ve lost that strength overtime by not finding the time to dance anymore. Muscles get stronger as we use them, and weaker when we don’t. The brain works similarly; it optimizes its resources to support frequent behaviors over behaviors it is rarely called on to perform (read more here if you’re interested in how). Since AI allows us to outsource tasks, there’s growing concern amongst researchers that we will exercise our own brains less and could lose proficiency at performing these tasks ourselves. 

This “use it or lose it” concern isn’t new, in Plato’s Phaedrus Socrates warns that writing will “create forgetfulness in the learners’ souls” because people will stop exercising their own memory once they can store things in writing and look it up later³. That was written around 370 BCE. Every technological revelation since, from the printing press to calculators to internet search engines, has triggered some version of this fear. 

These fears are not baseless. A 2011 study found that when people expect information to stay available online, they remember where to find information rather than the information itself. This has been coined as the “Google effect” on memory⁴. We get a clue for how the neural side of the story can play out by looking at the rise of GPS. Neuroscientist Eleanor Maguire had found that London taxi drivers, who needed to memorize roughly 25,000 streets to be licensed to drive London’s famous black cabs, had a larger posterior hippocampus (a brain region central to spatial memory) than non-drivers. The longer someone had driven a cab the bigger that region tended to be, just like a well exercised muscle⁵. Two decades later, McGill University neuroscientist Véronique Bohbot investigated what happened when technology entered the picture and found that people with greater GPS use over their lifetime had measurably worse spatial memory when navigating without it⁶. Their brain’s spatial memory muscle had grown weak from lack of use.

So far, there isn’t an equivalent study looking at the effects of long-term generative AI use on the brain and our cognitive skills, mainly because generative AI hasn’t been around long enough. There is work trying to understand if there are short term changes in brain activity while using generative AI. Early results suggest brain activity may decrease when offloading tasks to generative AI, but the existing data is limited and inconvlusive^7,10. One of the biggest open questions is how using generative AI may affect developing brains of kids and teenagers now who are growing up with LLMs. We don’t yet know how routine use of generative AI might shape attention, memory, or reasoning, especially if that use is happening while the brain is still “under construction”. We may be rapidly integrating the technology, but our understanding of what it means for our brains is not keeping up.

The case for friend: give our brain a break

While the use of AI to offload cognitively challenging tasks worries scientists, that functionality is also what makes it useful. As with any tool, its purpose is to make life easier. Outsourcing tasks to an LLM, which can likely do them faster, can free up our limited time and energy for rest, personal relationships, and everything else that is good for us and our mental health. Not having to expend mental capacity on tasks that an LLM can handle might also raise the ceiling of the work we can accomplish, much like a calculator helps us tackle more complex mathematical problems by taking the arithmetic off our plate. As with many tools, whether the widespread use of AI is “good” or “bad” likely depends less on the LLMs themselves and more on how we use it: as a collaborator we still think alongside, or as a replacement for thinking at all.

So, is generative AI a friend or foe for our brains and mental well-being? Neuroscience doesn’t know yet, and anyone claiming certainty either way is getting ahead of the data. What we have are early inklings: preliminary evidence for reduced neural engagement during AI-assisted tasks, a previously seen pattern that skills can fade when offloaded to technology, and no real idea yet what growing up alongside it does to the brain. One thing is for certain: the cat is out of the bag and it’s not going back in. Generative AI is here to stay, so these are concerns we can’t ignore and question we should be working to answer.

References 

1. Stanford Institute for Human-Centered Artificial Intelligence. “Economy.” The 2026 AI Index Report. Stanford University, 2026. https://hai.stanford.edu/ai-index/2026-ai-index-report/economy.
2. Ray, Siladitya. “Claude Makes More Money Per User Than Market Leader ChatGPT, Report Finds.” Forbes, 16 June 2026. https://www.forbes.com/sites/siladityaray/2026/06/16/claude-makes-more-money-per-user-than-market-leader-chatgpt-report-finds/.
3. Plato. Phaedrus. Translated by Benjamin Jowett, Project Gutenberg, 1 Nov. 2008. https://www.gutenberg.org/files/1636/1636-h/1636-h.htm.
4. Sparrow, Betsy, Jenny Liu, and Daniel M. Wegner. “Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips.” Science, vol. 333, no. 6043, 2011, pp. 776–778. 
5. Maguire, Eleanor A., David G. Gadian, Ingrid S. Johnsrude, et al. “Navigation-Related Structural Change in the Hippocampi of Taxi Drivers.” Proceedings of the National Academy of Sciences, vol. 97, no. 8, 2000, pp. 4398–4403. 
6. Dahmani, L., and Véronique D. Bohbot. “Habitual Use of GPS Negatively Impacts Spatial Memory During Self-Guided Navigation.” Scientific Reports, vol. 10, 2020, article 6310. 
7. Kosmyna, Nataliya, Emma Hauptmann, Y. T. Yuan, et al. “Your Brain on ChatGPT: Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing Tasks.” MIT Media Lab, preprint, 2025. 
8. Stankovic, Milos, Ella Hirche, Sarah Kollatzsch, and Julia Nadine Doetsch. “Comment on: ‘Your Brain on ChatGPT: Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing Tasks.'” arXiv, 2026, arXiv:2601.00856. https://doi.org/10.48550/arXiv.2601.00856.
9. Shaw, Steven D and Nave, Gideon, Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender (January 11, 2026). https://doi.org/10.31234/osf.io/yk25n_v1, The Wharton School Research Paper, SSRN: https://ssrn.com/abstract=6097646 or http://dx.doi.org/10.2139/ssrn.6097646
10. Stankovic, Milos, Ella Hirche, Sarah Kollatzsch, and Julia Nadine Doetsch. “Comment on: Your Brain on ChatGPT: Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing Tasks.” arXiv e-prints, Dec. 2025, arXiv:2601.00856. https://doi.org/10.48550/arXiv.2601.00856

Claude Sonnet 5 assisted with outlining and some wording. 

Cover photo from Immo Wegmann on Unsplash