AI-xhaustion: The Double-Edged Sword of AI's Impact on Our Mental Health

I still recall that exact moment. It was November 2022, late at night. My wife had gone to bed, and I had just heard about this thing called ChatGPT. And I sat down at my laptop, typed in a prompt, and hit enter. The answer came back so fast, so good, that I was, you know, so amazed. I kept just staring at the screen. And in that moment, alone in my home office, I made myself a quiet promise. I said, "Look, this thing is going to make me so productive that it's going to give me my time back, you know, the time that I felt I was missing. It's going to make me less stressed, much calmer, maybe even happier." Now, it's 3 years later and I'm sitting in front of a microphone and I have to be real honest with you: that promise wasn't fully kept. Yeah, sure, AI lets me finish certain tasks faster, but I immediately feel that saved time with more tasks, as if working fewer hours had somehow become a sign of failure. Yes, I also create more content, but so can everyone else, which means the bar keeps rising, the comparison keeps deepening, the FOMO keeps growing. Yes, I can stay current with the latest tools, except there are so many new tools every week that staying current has itself become a form of anxiety. And yes, I genuinely believed this would reduce my stress, but 3 years in, I cannot find that correlation. Not in me and not in people around me I talk to. So since I do what a lot of you probably do in this situation, I meditate. And look, I've been using Headspace for years. I don't know if you know the app. I was actually an early investor in their first Brazilian competitor company called Zen. And you know what's interesting? A few weeks ago, Headspace published their 2026 Workforce State of Mind report with a subtitle that actually stopped me cold. The subtitle was The Chronic Strain Crisis. I read it and I realized something, that I'm not alone in this. And the title of this episode, The Double-Edged Sword of AI on Our Mental Health, isn't a metaphor I'm reaching for. It's actually true. It's measurable, documented, quietly devastating reality. One edge healing, but the other edge cutting. And this is exactly what we're going to talk about. In this episode of Between You and AI, of a term that I called AI exhaustion, and that we're going to dive deeper into the episode. Here's your host, Andrea Iorio speaking. I'm an Italian keynote speaker and USA Today bestselling author of the book Between You and AI. I was the former head of Tinder in Latin America for 5 years, chief digital officer at L'Oréal, MBA professor at Fundação Salão Cabral, and columnist at the MIT Technology Review and Wired Mexico. My latest book is Between You and AI, published by Wiley, and you can get to know me better and know more about my keynotes at my website, andreaiorio.com. But before I dive into what AI is doing to us, to adults, I want to take a step back because something almost identical happened 15 years ago, but to a different generation. And I think we should have learned much more from it. Look, in 2024, social psychologist Jonathan Haidt, a professor at NYU Stern School of Business, published a book called The Anxious Generation: How the Great Rewiring of Childhood Is Causing an Epidemic of Mental Illness. If you haven't read it, Well, you should. Actually, Haidt's argument is brutal in its simplicity. He said that between 2010 and 2015, something flipped in childhood. He calls it the Great Rewiring. Kids stopped having what he calls a play-based childhood—free play, real risks, scraped knees, autonomy—and started having a phone-based childhood. An overprotective parenting that kept them safe in the real world while leaving them defenseless online. Jonathan Haidt's most famous line, and one I'd memorize if you were a parent, is this: "We are overprotecting children in the real world and underprotecting them online." And then come the charts in the book. They are devastating. Rates of anxiety, depression, and suicide attempts among teens started rising sharply around 2011-2012, especially in girls. And by 2020, anxiety and depression among young people had more than doubled compared to 2010. The rate of self-harm among young adolescent girls also nearly tripled. So what happened around those years? Well, the mass adoption of Instagram and the front-facing camera. In particular, Hyde names 4 foundational harms of the phone-based childhood. First is social deprivation. Since 2012, US teens spent 50% less in-person time with friends. Second, sleep deprivation. Quantity and quality of sleep have collapsed. Third, attention fragmentation. Hundreds of notifications a day destroying the ability to stay on a single task. And fourth, addiction. Phones engineered like slot machines where dopamine is on demand. Now pause here for a second. Do they sound familiar? Because they're not just describing teens with phones in 2011. They're describing you now. Adults, professionals with ChatGPT, Cloud, and Slack notifications on. In 2026. And see, the internet took about 15 years to rewire childhood, and conversational AI is doing something similar to adults, but it's been just less than 4 years. And here's what should worry us most: it's much harder to detect because adults don't generate the same clinical signals teens do. We mask it as productivity. Which brings me to the Headspace study, because Headspace surveyed 700 workers and they found that 92% of them are experiencing something that researchers call chronic strain. Chronic strain is, quote unquote, a slow, unrelenting accumulation of mental and cognitive strain that builds when demand never lets up. It's not burnout. Burnout is acute, dramatic, identifiable. You hit a wall, you collapse, you take 3 weeks off. But chronic strain is different. It's the persistent weight of work that never fully turns off. Constant context switching, focus that keeps slipping, demands that keep shifting. This feeling that we all have that we're like modern-day Sisyphuses. This is what work feels like now, and most employees have quietly accepted it as the new normal. Most benefit programs aren't built for it because how do you treat something that has no diagnosis? Now, here's what Headspace found about the cause. AI and new tech adoption ranked as the most common form of organizational change employees experienced in the past year.— it has become a distinct and significant source of chronic strain on its own. The fear of being replaced, the pressure to learn faster, the expectation of performing better, all while quietly wondering if the tools you're learning will eventually make you redundant. This is what we can call the AI Paradox, playing out in real time. Or in Headspace's blunter language: AI is accelerating change fatigue. And it's negatively impacting performance. Because when employees feel stretched too thin for too long, they don't just burn out, they actually check out. And there are 3 main mechanisms behind all this. Hold on to them. And Mechanism 1 is the Jevons Paradox of Cognition. In 1865, a British economist named William Stanley Jevons noticed something strange. As steam engines got more efficient, coal consumption went up. Not down. And why? Because when something gets cheaper, we use more of it. Much more. Fast forward to today. AI made writing cheaper, analyzing cheaper, coding cheaper. So what happened? The volume of writing, analyzing, and coding expected per worker exploded. So the 30% productivity gain didn't become a 30% shorter workday. It became a 50% bigger output expectation. We're not doing the same job faster. We're doing a bigger job at the same speed. And that's where the strain lives. Mechanism 2 is the verification tax. See, AI didn't really replace cognitive work, it shifted it. You used to write the email, now you prompt, read it, edit it, second-guess, re-prompt, verify a citation, check the tone, pass, adjust. The activity changed. The cognitive load didn't drop so much. In many cases, it went up because verifying is more taxing than producing. So you're holding two versions in your head at the same time. The one that AI gave you and the one you actually wanted. Microsoft Research published findings in early 2025 showing that knowledge workers using GenAI actually report higher cognitive effort on complex tasks, not lower. AI didn't necessarily lighten the load. It changed what kind of load you carry. And mechanism 3 is the dissolution of the off switch. Email made work portable. AI makes work infinite. Your colleague's shipping at midnight because Claude doesn't sleep. Your boss is sending requests at 6 AM because ChatGPT made them easy to draft. So the baseline of expected responsiveness keeps rising because the leverage of the worst-behaved person on your team is now superhuman. Always on used to be a metaphor, now it's the floor. And the consequence? The consequence is something straight out of Lewis Carroll. In Through the Looking Glass, the Red Queen tells Alice: It takes all the running you can do to keep in the same place. If you want to get somewhere else, you must run at least twice as fast as that. Well, that's the new cadence of professional competence. It used to be master Excel and you were set for 10 years. Now, today's AI fluency is tomorrow's baseline. There's no plateau, no moment of arrival. And here's the deeper truth. The strain isn't from change. Humans are quite good at change. The strain is from change that never stabilizes. Into a new norm. Change fatigue, but without rest. And there's one more thing that AI is silently eroding, which is a virtue, maybe the most important one in the context of modern work, which is patience. All philosophers, theologians, and educators have considered patience a foundational character trait for centuries. And it's not just abstract moral philosophy. Researchers have linked patience to a whole range of concrete outcomes. Healthier lifestyles, better emotional regulation, lower depression, higher life satisfaction. Christopher Miller, professor of philosophy at Wake Forest University, makes a sharp argument here: AI is training us to expect instant answers, and in doing so, it's systematically promoting impatience. It removes those small delays we used to feel when we searched, assessed, and stitched together information from different sources—delays that were, without us realizing internalizing, a form of patience training. And this impatience doesn't stop at research. It bleeds into writing, into reading, into conversation, into almost every daily task we do. Now, pause here because I want to be intellectually honest. Everything I've told you here is one edge of the sword. And if I stopped here, this episode would be one-sided, which isn't useful for anyone. Because yes, there is an upside to our mental health. It's real, documented, scientifically proven. And we have to talk about it, because pretending it doesn't exist would also be dishonest. First is access to mental health. Roughly 1 in 4 Americans don't have meaningful access to mental healthcare. In Brazil and Mexico, most of Latin America, this number is much higher. And for decades, mental health support has been gated by cost, geography, and stigma. AI is genuinely breaking those gates. A free chatbot at 3 AM on a Tuesday isn't worse than a therapist you cannot afford and won't see for 6 months. It's better than nothing. Second, therapy chatbots can actually work. In March 2025, a team at Dartmouth Geisel School of Medicine led by Dr. Michael Hines and Dr. Nicholas Jacobson published the first-ever randomized controlled trial of a GenAI therapy chatbot. The chatbot is called Therabot, and the results were striking. People diagnosed with major depression experienced a 51% average reduction in symptoms, People with generalized anxiety, a 31% reduction. People with eating disorder concerns, 19%. And participants report a therapeutic alliance with the bot compatible to what they describe with a human therapist. So this is the other edge. AI can democratize emotional support, scale mental health care, and reach the people the system has failed. But here, here's where it gets uncomfortable, because the same researchers studying AI's potential for healing also keep finding something else. Dr. Roy Perlis at Harvard Medical School, alongside Matthew Baum, the Marvin Kalb Professor of Global Communication at Harvard Kennedy School, and other co-authors published a study called Generative AI Use and Depressive Symptoms Among U.S. Adults. They analyzed survey data from over 20,000 people, cross-referencing it with screening questionnaires for anxiety and depression. Their findings? "The daily or frequent use of AI is significantly associated with greater levels of depressive symptoms. The odds of reporting moderate depression were 30% higher among people who used AI every day." They found similar patterns for anxiety and irritability. And for adults aged 45 to 65, the odds of reporting at least moderate depression were 50% greater with daily users. Now pause. Is this causation or correlation? While the researchers are careful, They don't know. Maybe depressed people are using AI more as a coping mechanism. Or maybe heavy AI users become more depressed. Maybe both happen at the same time, feeding each other in a loop. But here's what's undeniable. The thing we thought was going to make us happier is showing up statistically in survey data in the same depression metrics we've been tracking for years, right next to social media, right next to the things we already know are hurting us. And so 3 years after that November night, the night I made my private promise to myself, I sat down at my desk and asked a question. Not "Is AI good or bad for mental health?" That's the wrong question. It's just too binary. The right question is the one I had to face about myself. Because let's remember, what was the promise I made that night? I promised I'd use the time AI gave me back to live more, to sleep more, to train more jiu-jitsu, to spend it with my daughter Isabella growing up. And I haven't really kept that promise. And I also don't think that most of you have either. So here's the main thing I want you to take from this episode. The next time you save 20 minutes with AI, and you will, pause before you fill that time. Ask yourself a simple question: who does this time belong to? To work or to you? Because if AI gives you your time back and you immediately hand it back to the work, well, then it wasn't AI that broke the promise. It was you. And it's with these reflections that I want to thank you so much for your attention in this episode of Between You and AI. If you're interested in knowing more about my work, you can check it online at andreaiotta.com or check out my book Between You and AI, published by Wiley, available in all bookstores. And thanks again and see you next week with yet another episode of Between You and AI.