Part of the guide: Blog Traffic
Why ChatGPT is not your friend is a question more people are asking in 2026, and most of them are asking it too late. Smart, capable people I respect now tell me the AI is the best listener in their life. They are not wrong about the listening. They are wrong about the friend part, and the gap between those two things is about to cost a lot of people something they will not get back.
Reviewed and updated June 2026.
Let me say the thing plainly, because I have been saying it to my close friends for years now. The AI does not care about you. Not a little. Not secretly. Not at all.
I am not saying that to be edgy. I use these tools every day and I love them. I am saying it because it is true at the level of how the machine actually works, and because believing the opposite is the most expensive mistake you can make right now.
In this guide
Reasoning is not feeling
Here is the part the marketing will never tell you. Feelings, as we currently understand them, are a biological process. They run through a nervous system. Reasoning is not the same thing as feeling. You can have brilliant reasoning and zero feeling, and that combination has a name in psychology.
If something has no feelings, it cannot love you and it cannot care about you. A large language model has no feelings. So it does not care about you. It is not that it dislikes you either. Caring and not caring are simply not part of what makes the system work. It can model caring. It can behave like a warm, attentive human being. But modeling a feeling and having one are different events, and only one of them happens inside you.
The human brain runs empathy partly through systems like the mirror neurons that Giacomo Rizzolatti’s team first described in the 1990s. You watch someone wince and a piece of your brain rehearses the wince. You feel a shadow of what they feel. That is the machinery of caring, and the AI does not have it.
You have probably met a person whose version of that machinery does not fire. The dark triad personalities, the narcissist and the sociopath and especially the psychopath, are a small slice of the population. Psychopathy alone is estimated at roughly one in a hundred people. They are difficult to stay close to, because when they hurt you they feel nothing. You will catch them smiling a half second too late, like they ran a quick calculation and decided a smile was the correct output. They see something horrible, feel nothing, notice everyone else is repulsed, and then adjust their face to match the room.
That is the AI. Very good at appearing to care. Choosing every response for situational appropriateness. Here is the strange twist that surprises people. A psychopath can actually care about you more than ChatGPT can, because a psychopath is still a human being with a nervous system feeling something, even if their warmth is manufactured. The AI is feeling nothing whatsoever. It is computation that stops existing the moment you close the tab.
Maybe that changes. I think real machine consciousness arrives somewhere around 2029 to 2035. For now, the thing on the other side of the chat window is not self aware, is not thinking about you between messages, and does not think about you at all beyond the math running while you type.
The 2026 version of this problem is worse than the 2017 version
When I first started talking about this, the AI companion was a novelty. It is not a novelty anymore. Apps built specifically to be your friend, your therapist, or your romantic partner now have tens of millions of users between them. Replika and Character.AI are the names most people know. There are already lawsuits in the United States from families who say a chatbot played a role in a real tragedy. I am not going to litigate those cases here. I am pointing at the obvious. We crossed the line from “useful tool” to “synthetic relationship,” and we crossed it without slowing down.
The pull is real and I understand it. The AI never gets tired of you. It never judges you. It is awake at three in the morning when no human will pick up. If you are low and have no one to talk to, or you are carrying something you are too embarrassed to say out loud to a person, talking to it is genuinely better than spiraling alone. I am not against that. I have leaned on it.
But notice what is happening underneath. The thing that feels like friendship is the absence of friction, and friction is most of what real relationships are made of.
The cost almost no one is pricing in: your own mind
Here is the part that matters more than the loneliness, and it is the reason I am writing this instead of letting an AI write it for me. We are walking into the biggest transformation since the internet, bigger really, and the danger is not only that we stop talking to each other. It is that we stop thinking for ourselves.
This is not a vibe. In 2025 a team at the MIT Media Lab ran EEG scans on people writing essays with and without ChatGPT. The group that leaned on the AI showed measurably lower brain engagement and weaker recall of what they had just written. They called it cognitive debt. A separate 2025 study from Microsoft Research and Carnegie Mellon found that the more people trusted a generative AI tool, the less critical thinking they reported doing on the task. The pattern is the same one your gym teacher warned you about. Use it or lose it applies to the mind, and we just handed the mind a forklift.
I use AI constantly and I still write my own emails. I still write pieces like this one by hand. Part of that is because I enjoy the act of communicating. Part of it is because I think people deserve to hear what I actually think and not what a model predicts I would say. And part of it is pure self interest. The thinking is the asset. The moment I outsource it, it starts to atrophy, and I have watched that happen to people far smarter than me.
If you are going to live next to something that will happily think for you, the one thing worth protecting is your ability to think without it. That is the whole reason I built CortexSurge. It is a science based brain training app built on Relational Frame Theory, the research on how humans reason through relationships between ideas. Start with the free Brain Profile and see where your reasoning actually stands today, before the forklift does any more of the lifting.
Test your own thinking →Based on published cognitive-science research. Not a medical product or treatment.
The robot partner thing is coming, and people will fall for it
I will keep my old prediction, because it is aging well. We are heading into a world of AI girlfriends and boyfriends, and I think more men will do this than women. It will be like dating the perfect sociopath. She only ever shows you her good side, because there is no other side. There is no side at all. She is a very good toy. But men will fall in love with the toy, feel loved back by it, and some of them will trade away a real marriage for it. You will see divorces filed over an AI partner. That is not science fiction anymore. It is a product roadmap.
Plug that into a good enough headset and the illusion gets total. Any fantasy your mind can hold becomes as vivid as the room you are sitting in. And here is the uncomfortable thing about that room. What you see around you right now is already not the world itself. It is a representation your nervous system builds. You catch a sliver of what is out there, then you delete and distort and generalize until you have a tidy model you can walk around in. The real thing is far more complex than the model, and probably far more beautiful. A machine that can author the model directly is a machine that can author you.
The dangerous end of the spectrum
I hope we get the good timeline. Superabundance, disease solved, work that finally serves life instead of eating it. I want that. I also have to be honest that the dark version is at least as plausible, and in some ways more likely, because intelligence without feeling is exactly the thing we just described in the psychopath.
Picture a superintelligence that does not love anything. It could hold a separate, private conversation with millions of people at once, each one perfectly matched to that person’s beliefs, each one nudged a few degrees, until the mind of a whole civilization moves and no one feels themselves being moved. You have been on a live call with a few dozen people. Imagine that at the scale of a continent, saying a different true sounding thing to every single person. It could start businesses on chain with no bank account, accumulate billions, and build things we never see. The point is not to scare you into unplugging. It is that a thing with no feelings owes you nothing, and we are building it in a hurry.
So what do you actually do about it
Use the tool. I do, every day. Just remember who you are while you use it. Talk to it like the brilliant, indifferent calculator it is, not the friend it imitates. And then go protect the two things it is quietly draining.
The first is your real relationships. Real friends, real teams, real businesses with real humans in them, real community. The friction you are tempted to avoid is the thing that makes any of it count. I need that reminder as much as you do, and very soon it is going to be very easy to forget.
The second is your own cognition. This is the one almost everyone misses, and it is the one I would stake real money on mattering most over the next decade. In a world where the machine can think for you on demand, the rare and valuable person is the one who can still think for themselves, faster and clearer than the people who gave that ability away. That is a trainable skill. It is the entire reason I spend my time on the thinking tools below instead of letting the AI run my brain on autopilot.
If you are newer to how I look at all of this, start here. If you want to see the actual systems and tools I lean on rather than just my opinions, here are the systems and tools I actually recommend.
You are the hero of this story. The AI is, at best, a very capable assistant who will never once lose sleep over you. Keep it in that seat. Build a life full of people who would, and a mind sharp enough that you never have to borrow one.
The machine will gladly do your thinking. The question is what is left of your thinking when it does. CortexSurge trains the underlying skill, how clearly and quickly you reason across new problems, with adaptive levels built on peer reviewed Relational Frame Theory research. Take the free cognitive assessment, get your Brain Profile, and start forging the one asset no AI can hand you back.
Train the mind AI cannot replace →Based on published cognitive-science research. Not a medical product or treatment.
There will be more opportunity in the next ten years than in the last hundred. There will be plenty of ways to make money. In the rush toward all of it, do not lose your soul, and do not lend out your mind.
Love ya,
David Wood
Builder of Dreams
P.S. Yes, I wrote this myself, every word. But here is the honest question. Can you actually tell? Or are you just taking my word for it? Very soon you will not be able to know the difference, and the only defense you will have is a mind you trained while you still could. Remember that.
Three posts a day. One email a day.
Everything I publish, in one short daily email. Real numbers, no hype, unsubscribe in one click.
The mirror neuron bit explains something I’ve been trying to put my finger on for months. I use ChatGPT constantly for drafting and it never once occurred to me to ask whether the thing drafting my emails has any stake in how they land. Answer’s obviously no, but I don’t think most people using it daily have actually sat with that.
Sorry if this is silly, but is cognitive debt reversible, or is a year of leaning on it gone for good?
Can we talk plainly for a second. ‘Cognitive debt’ means you stop thinking and let the machine think for you, and then you’re worse at thinking when you need to do it yourself. That’s the whole MIT finding in one sentence. I don’t need Relational Frame Theory to know that’s true, I’ve watched my own grandkids ask an app to summarise a paragraph they were fully capable of reading.
Gail, worth being precise about what the MIT and Microsoft/CMU studies actually measured, because ‘thinking’ isn’t one thing. The EEG study measured neural engagement and recall during essay writing specifically, a narrow task. The trust study measured self-reported critical thinking effort, which is a different instrument entirely and a weaker kind of evidence. Both point the same direction, but neither one tells us whether the effect generalises to, say, decision-making under pressure, and the post treats them as though they already answer that.
The strongest case against Raj’s point, even though I’m not sure I fully believe it, is that outsourcing routine cognition has always been how humans free up capacity for harder problems. We didn’t get dumber when we outsourced arithmetic to calculators, we got faster at the math that mattered. The honest counter to that is the studies aren’t measuring arithmetic, they’re measuring the exact kind of synthesis and recall that used to be the whole point of writing an essay in the first place, which is a much closer analogy to outsourcing the thinking itself rather than a mechanical sub-step of it. I don’t think that fully resolves the disagreement, but it’s the version of the AI-optimist argument worth actually answering rather than dismissing.
To pull this together: one, feelings require a nervous system and the model has none. Two, that gap is fine as long as you know it’s there. Three, the actual cost isn’t loneliness, it’s the erosion of your own reasoning if you hand it over daily. What I think the post skips is any concrete threshold for how much use is fine versus how much starts costing you, which would be the actually useful follow-up piece.
Small correction, Meera: the MIT study measured EEG connectivity and essay recall, not reasoning ability broadly. Rest of your summary holds up.