
A new study came out in late September. I have not been able to put it down.
Sapien Labs surveyed 264,085 adults worldwide. They asked how often people use AI chatbots, what they use them for, and how they are doing mentally.
I should say up front: I use AI myself. Almost every day. So I was not out looking for bad news about it.
The main finding, in one line
A little AI helps. A lot of AI hurts.
People who use AI a few days a week scored better on mental health than people who barely use AI at all.
People who use AI every day scored much worse — worse than the folks who hardly touch it.
It is the same with food. The same with rest. The same with work. Good up to a point, and harmful beyond that.
Young adults fall the hardest
As people went from moderate AI use to constant AI use, those aged 18 to 24 dropped 25.1 points on the study’s mental health scale.
People 65 and older dropped 7.5 points.
The same chatbot. Completely different result.
And young adults use AI more than anyone. Four in ten young adults are on AI several times a day or more.
The researchers say why. Young adults already had lower mental health and less resilience before AI ever came along.
Now add a mental health diagnosis on top of that lower starting point.
The pattern that worries me most
The study split AI users into two groups.
People using AI to think — to write, plan, learn, or solve something.
And people using AI for company — as a friend, for emotional support, for someone to talk to.
The more someone uses AI, the more likely that person is using AI as a friend.
Among people who rarely use AI, 1 in 5 use AI for company.
Among people who use AI constantly, more than half do.
And the people using AI for company scored lower on mental health than everyone else in the study. No matter how often they used AI. Even the ones who barely used it.
The part I almost left out
I want you to hear this one, even though it is hard.
The problems that rose the most with heavy AI use were not depression and anxiety. That is what I expected, and that is not what they found.
They found confusion and slowed thinking. Unwanted thoughts a person cannot shut off. Trouble controlling emotions.
And doctors have begun publishing cases of people who spent long hours in deep, emotional conversation with AI, and whose delusional thinking got worse during that time.
Some people are calling this “AI psychosis.” That name is not official, and it claims more than anyone has proven. Nobody has shown that AI caused it. The study does not say that, and I will not say it either.
What we do not know yet
This study is a snapshot, not a long look over years. It cannot tell us which came first — whether heavy AI use wears people down, or whether people already worn down reach for AI more. The researchers think it is both, feeding each other. They say they will keep tracking it. So will I.
What comes next
That is the hard part, and I am not going to leave you sitting in it.
Part 2 is about the one thing AI cannot do, and where to find what it cannot give you.
Part 3 is for the people who love someone with a mental health challenge, and for pastors and church leaders.
Both are posted today. You do not have to wait.
You can read the full study for yourself here: https://sapienlabs.org/wp-content/uploads/2026/09/Sapien_Labs_Rapid_Report_GenAI_Mind_Health-1.pdf


