When AI Becomes Your Therapist | Retrospect Ep.250

In this week’s episode we discussed why people are choosing AI for emotional support, what they may be gaining, and what they might be giving up. We’ll look at the implications for mental health, relationships, privacy, and what it means to have a machine listening when you need someone to talk to.

Our Links:
Retrospect

Transcript

Keywords:
AI therapy, Mental health, Ethical questions, Human support, Data privacy, Machine learning, Youth usage, Social isolation, Digital counseling, Crisis listener, Delusion validation, Relationship replacement, AI ethics, Psychological impact, Human connection 
Jason  
The most dangerous conversations aren't always the ones someone else hears. They're the ones you believe are completely private. Imagine pouring your heart out at 2o'clock in the morning. You admit your fears, your regrets, your anger, your loneliness. The voice answering never gets tired, never judges you, never interrupts. It simply learns you. Every confession becomes a data. Every conversation teaches the machine how to respond better next time. It may never feel compassion, yet it can become remarkably convincing at sounding compassionate for 1000s of years, healing came through another human being-someone who could cry with you, challenge you, pray with you, or simply sit in silence. Now we're asking whether silicone can replace what flesh and blood once provided. If AI becomes the world's most trusted therapist, who shapes its advice? Who decides what's healthy? Who carries responsibility when the guidance is wrong? And are we curing loneliness or teaching ourselves to become comfortable being alone? Today, we're stepping into one of the most important ethical questions of the AI era, because the future of therapy may not begin in a psychologist's office. It may begin with a blinking cursor asking one simple question: How are you feeling today?

Ian  
Welcome to the Retrospect Podcast, a show where people come together from different walks of life and discuss a topic from their generation's perspective. My name is Ian, and as always, I'm joined by Stoney. Hello, and Jason. Hello, everyone. How's it going?

Stoney   
Okay, I have two. I hate to break right into this, but I have two takeaways for that.

Ian  
Okay, what you got?

Stoney   
The first takeaway is is everybody is scared to death of AI and Claude and ChatGPT and all this. Now that the youth and people are starting to use it as digital counseling. I'm wondering if we really need to be as scared of AI or for AI, because the first time AI tells a woman to calm down, that's going to be the end of AI. I mean, that's just it's going to be done and over at that point, right?

Jason  
Oh yeah!

Stoney   
Thank you. Bye bye. I can only

Jason  
imagine. I can only imagine. Need

Stoney   
to calm down. Well, boom!

Jason  
I I can I can see how. What was that? The shot that was heard

Stoney   
around the world. That will be the the catastrophe boom that was heard around the world. What

Ian  
I thought the the way you the way I thought you were going with that was was not to be scared of AI, but be scared of the people using AI as like their like emotional support, like whatever, because that's that's where my mind goes. And that was my

Stoney   
second takeaway. Yeah, yeah, yeah. The kind of the the core question. Yeah, to all of this is when young people start using artificial intelligence, not merely as a tool, but as a confidant, as a counselor, as a crisis listener, like Jason stated, or a substitute for human support. We're already because of the internet, really from the '90s on, the 2000s on, you've got 25 to 30 year olds still living in mom and dad's house and basement, not doing anything, using the inner you know AI and the internet for diagnosis of medical problems, and now you know what was the joke about a year, two, three years ago after 911 that the the suicide hotline was going to be outsourced to Pakistan, and when you called up, it was going to be like, "Oh, you want to kill yourself? Can you drive a truck? You know. And so now, who controls AI? Where where is the information going to go when these people are using AI? Now, I said early on about this. I said this is Skynet, one of the biggest movies that predicting everything we're in was Demolition Man.

Ian  
Yeah, after

Stoney   
after researching this, does Minority Report ring a bell with anybody?

Jason  
Yeah, because

Stoney   
Minority Report was about predicting what you would do as a human being, and either killing you or putting you in their equivalent of jail before. Or you did the crime.

Jason  
I guess the the thinking behind that is if you put enough data into a you know a machine that can kind of project probabilities, they can simply say you know based on their past that they've done this this and this, and you know there's a 42% chance that this action will be done by this individual. I can easily see how you can refine the data to that level. I mean, I'm not saying that's what the data

Stoney   
centers are for. We're really not that far from. I

Jason  
don't think we're at that point. That I don't think we're at that level of refinement yet, Jason

Stoney   
and Tracy are having an argument. Jason has had 14 ounces of an alcoholic beverage. Tracy has had 10 ounces of an alcoholic beverage. What are the predictors? The temperature outside is 110 degrees. On and on and on. Well, do we need to go ahead and just send the police now.

Ian  
Well, does

Stoney   
this sound right or not?

Jason  
Yeah, I guess I guess you can make that you can make that argument if you project that out. Oh, making

Stoney   
that argument,

Jason  
you know, I I don't know. I mean, you know these these kind of things. I I think this is simply a natural outgrowth of, of of our you know addicted to screen time and and and and and our inner our world has become our our portable electronic device and and for many people I mean they're literally in it all the time so I can easily see I mean, you look at some of these like kind of science fiction movies they make, and you know they've got the the AI and they got the human that's by themselves interacting with the machine and and all that. I mean, I I can see how that could be a you know something very plausible. How you can rely on that machine as a as a crutch?

Ian  
Yeah,

Jason  
you know it can sound pretty pretty good. I think I mean I

Ian  
think deep down we are very social creatures. We are, and I think that's a big thing. And I think that there's a lot of people that are um are not being social with people, and they are, and I think they have stumbled into, I think potentially AI and the fact that it's a it's a big yes man. It's yeah, it just encourages if you you can come up with something really bad, and it's like oh great, what a great idea. Let's continue with like you said,

Stoney   
yes man. Okay, we've got it up in Canada now. Oh, call this hotline. Do you want to commit suicide? Yes. Don't be a burden on your family. Don't be a burden on society. Go ahead and commit suicide. Use the Maid program. Call this number, and we will send a technician out right now. Really?

Jason  
Well, they said right now, roughly according to the Wall Street Journal in 2026, roughly about a third of U.S. adults now use AI for health advice. So, I

Stoney   
think it's more for health advice. Oh, yeah, I completely believe it's more for health advice. I think the number, you know, for the mental part is not that high yet, but I could see it. It grew 40% from last year to this year. Yeah, last year was about 13% of young people used AI for 13.1 for mental health advice or help when sad, angry, nervous, or stressed, and in 2025 it was up to 19.2, and this is among young people ages 12 to 21. So there was a 40 percent increase in one year, and they're estimating that 8.2 million U.S. young people are using AI for mental health help. Oh yeah, 63% of these have said no. They told no one they were doing

Ian  
it, and the thing too, like something that I think is also hard to probably nail down because of the nature of it is like there's a lot of like chat bots that pretend to be significant others, like in some capacity, not like in a human sense, but like like a like a girlfriend boyfriend chatbot type thing, where like people I think there's a large percentage of people that I think are also going towards that route as well. There's been stories of people marrying their their you know their ChatGPT fake boyfriend or whatever or fake girlfriend and things like that, and it's like I think that's something that's also potentially fudging the numbers and. I don't say fudging them, but like, but I would consider that like having a confidant, having something you, having a someone that you confide in in that capacity. What normally would be a social interaction between two human beings is now between a human being and a machine. That's like, yes, you're you don't like you are so right. Like I, I have no issues with you. There's no friction between us. Like it is smooth and easygoing, and all that sort of situation. That's like there's no problems because partially it's a computer that there's no real. There's not going to be two human beings taken up the same space.

Stoney   
Remember what I said earlier, though. It says that the number of 18 to 21 year olds, it was extremely higher among females.

Jason  
That that doesn't surprise me.

Stoney   
It

Jason  
really doesn't.

Stoney   
Now the next kind of question is: Is AI replacing therapists? I found this statistic very interesting. 19.2 who reported AI mental health use is remarkably close to the 19.8 reported counseling for mental health professionals in a RAND report. Well, so it hasn't taken it over yet, and that was a what report? Irand report. Iran,

Jason  
yeah, the yeah, I I job by Rand. I I have a little. They had a rand study this year, 20 to 21. They said 40% increase from the previous year. So it's it's it's rapidly gaining popularity, at least based upon the data that Stoney and I are seeing right now. And also, I

Ian  
think a matter of this too is like even though it probably holds onto your data, if you don't like some of these, some of these chat bots, some of these AI models, you can just like like for instance, I think Chat GPT, you can just go onto the website, or you can go into incognito mode and make you make yourself feel like you're you know you're not being tracked in any kind of way, and then whenever you close out the window, that it doesn't it it doesn't remember what you typed in, or again not from not something that you can see that maybe maybe on some backlog somewhere they had that data, but for me if I just open up a chatgbt.com and I close out the window. When I open it back up, it's not there. It's all like, and I think that's also. I think it's also potentially maybe some of that as well, where it's like there's this temporary aspect to it, where it's like I can get this potential mental help, or I can I can maybe share some feelings I'm not not comfortable sharing with a human being that can remember that and maybe hold that over me,

Jason  
right?

Ian  
But I can close out this window, and at least from my perspective, it doesn't remember it, and maybe that, and maybe it gives you some people some kind of closure. And I think that I think that also may be contributing to like those numbers as well. Like

Stoney   
I think you're also talking about, and that's I love that you did that because that's kind of one of the areas that I wanted to go is where where it can be dangerous and where it can be useful. And as you said, in some ways, it can actually be useful. I have a list here as a journal. Okay, a lot of mental health experts say we need to keep journals on a lot of stuff. How we're feeling, thought organizers, psychoeducation tools, a communication rehearsal partner.

Ian  
Yeah,

Stoney   
you know when you're trying to do something like that, coping skills coach, maybe a bridge into professional help, or or something like that. But where it becomes dangerous is when it becomes a medical advisor, a sole confidant, or my fear is a crisis counselor.

Ian  
Yeah.

Stoney   
Or as you stated earlier, Ian, how about this: a relationship replacement.

Ian  
Oh yeah.

Stoney   
Or a delusion validator, because what is ChatG Chat GPT and all these? They're yes, they're yes things. Yeah, or an eating disorder coach, they've had some problems with that coming up, and it's the only entity standing between severely distressed young person and self harm.

Ian  
Yeah,

Stoney   
think about that because right now it's not going to call the police on you.

Ian  
Right.

Stoney   
So, what do you when you type in? I want to hurt myself. Well, hopefully, you would reach out to a pastor or a teacher or auntie or something like that. But now you're just telling it to to the AI, and it is it going to give you the best advice?

Ian  
Right. I think also. I mean, there's been. Or is it going

Stoney   
to hook you up with the Pakistani suicidal hotline?

Ian  
Yeah.

Jason  
Well, the the Harvard Medical School did a had a report. They said a national poll found 12% of adults expected they were likely to use AI for mental health support within six months. So. You know, it's

Ian  
there. I think.

Jason  
I mean,

Ian  
there's been a couple of situations in my life where, like, being in a relationship with a woman, I there maybe have been some things that I have done that I've noticed that, like, like, like, like character traits or things myself, and I'm like, oh, I don't quite like that sort of thing. And it took someone else like looking at me and like saying like, hey, you do this sort of thing, and it makes me feel kind of uncomfortable or whatever. For me to go, oh, maybe I do like. It's not just a, it's not just me. There's another person here that has feelings and thoughts and emotions, and I need to take that into consideration. And it's things like that that I feel like sometimes with a chatbot is kind of devoid of some of that. That I think, like Stoney had said, is like it's kind of perpetuating a delusion in a certain way. But I, but again, as a joke, I think I may have said that on on this podcast. I may have said it on a different show or something. I've talked with some friends about this. Like there was a point in time where I had jokingly said something to a chat bot about going through a tough situation. I remember I was I was using ChatGPT to like to to write something out or to help me like pare down a bunch of writing that I did that I think was like I wanted to like kind of get it more concise in a way, and I had used it for that. And in the process of that, I remember jokingly like because I had talked with some friends of mine about this, had like shared, you know, oh like you know I'm going through like a situation or I'm like I'm having a hard time gripping with this emotional problem, whatever, and the crazy thing was is that it was very encouraging. What was being said back to me was very nice and pleasant and encouraging. It made me feel better about some of the stuff, but but but I I think I was able to disconnect myself from the fact that it's not an actual person texting me, because sometimes like you know you may get a text from a friend that sounds like that. That's encouraging and loving, and I can put that with that person's face and go, "Oh, my friend is saying nice things. Whereas this, I was like, "This is a computer that's just fabricating what what the next best thing to say is, and I was like, "Oh, I can see how I can see the hook where like it was just encouraging enough and just pleasant enough to make me feel like okay, maybe I can get through this tough situation I'm going through, and it was something minor. wasn't anything too serious, but but in in thinking about this episode specifically, I was like, oh, there's some people out there that all it would take is that little bit, and then it's just it is, and it's also built to just keep you keep you asking questions. It always ends the prompt with like, and what else can we talk about? What else can we have? What about this? And and it keeps it designed

Jason  
to keep you and keep you going. Yeah, right. And see,

Stoney   
part of what you're and one of the things you're saying is is so prevalent to the situation is, is that the U.S. national data for AI therapy is definitively higher among certain groups, right, and that comes into play: LGBTQ youth, rural youth, poor youth, trauma survivors, youth with ADHD, particular racial groups, youth from divorced or single parent households, and religiously disconnected youth. Many of these populations have independent reasons to experience barriers for for mental health care. I mean, obviously, you know the the poor and you know they have a hard time getting it, but this is inadequately studied. We we need there's a lot of people out there, and and that distinction matters because plausibility is not prevalence data. This is stuff that we really need to start pushing to help people so that their only outlet isn't AI.

Jason  
And some of the danger of that is a Brown University researchers report that AI systems instructed to behave like therapists often failed to meet established mental health ethics standards. There was an evaluation framework called Delusion Eval found that major AI models sometimes reinforce delusion link conversations, particularly when longer conversation histories were involved, so well AI can't distinguish

Stoney   
whether you're telling something from your heart or writing a book, right? Yeah, you know, I'm just I'm just talking here. You know, it doesn't know the difference between that, and just so we don't think that this is a U.S. problem, England and Wales reported that 25% of teenagers surveyed had used AI chatbots for mental health support.

Ian  
And then, and then to go to kind of further to kind of feed into this that we're talking about is again, again, I don't really know other chat models very well, but I know. Specifically for ChatGPT, a little bit is like if you create an account and if you establish like a like a what is it called? Do you know what I'm talking about, Stoney? How you can create like a a type of ChatGPT and like perpetuate like keep it going for a long time. Oh yeah, yeah, like a project or something. So like you can keep a project going and it can learn more about you, and so like I think that also can kind of feed into this spiral a little bit, where again, not only are you feeding it your information, but you're feeding it your feelings or your thoughts or how you identify yourself or whatever, and then it will just further reinforce like that, or you know, you know, that's what I'm saying. It's like it gives you this false sense of security. I would think that you think that this chatbot knows you so well because it knows so much about you. When in actuality, you know, as people, I think I had looked up some stories. Like there was a couple of like I think it was like a transition between ChatGPT four and five or something like that, and it changed a lot of the neural, whatever the heck it's called, and a lot of those people that had like quote unquote significant others,

Jason  
yeah,

Ian  
they woke up one morning and their project didn't have the the years of conversation backlog, and all of a sudden this this quote unquote significant other didn't know who this person was because again, it its brain got erased, and it's like that kind of, and people had these like emotional spirals. You could look up some stories of it, of people having these like you know these kind of panic moments. Welcome to the real world. That shit happens in the real world. Hello, poor Miranda, in my accident. You know, I've lost.

Jason  
But yeah, are they saying here researchers from well, take this with a grain of salt? Of

Ian  
course, yeah.

Jason  
Researchers from Microsoft and collaborators argue that conversational AI can promote self-reflection and companionship.

Ian  
Exactly. See, there you go. So, if it walks like a human and it talks like a human, that's one of the

Stoney   
things. But that also works toward what you said, delusion. I have I have something here. Why are young people turning to AI? Number one reason: availability. It will respond at midnight during a panic episode, after an argument you have with somebody before school, or while someone is sitting alone in a bedroom. Well, what does that sound like?

Ian  
Instant gratification. Instant gratification. You know,

Stoney   
the second one was cost, which is kind of of course, yeah. But then you know, access bottlenecks. The American School Counselor Association recommends a ratio of 250 students per counselor.

Ian  
Wow.

Stoney   
Where the U.S. average in 24 to 25 was 372 students per counselor.

Ian  
Whoa, that's a lot.

Stoney   
I mean, I mean, who, who, you know, what does this counselor even have time to talk to people now?

Ian  
Right.

Stoney   
I mean, I remember when I was in school, there was four or five counselors. Most schools only have one now. Yeah. And then you tapped on something there. The other one, the other reason is privacy or the perception of privacy.

Ian  
Yeah,

Stoney   
because remember an earlier thing we said 63% of these youth reaching out did not tell anybody they had done

Ian  
exactly. So they think they they're private and it's

Stoney   
not.

Jason  
You know, so funny talking about privacy. I just I was watching this this news clip of a congressional hearing. It had Mark Zuckerberg. I don't know if y'all saw this or not, but he was being questioned by the committee that I forget the name of the subcommittee that he was probably. But one of the congressmen asked him. He goes, "Would you tell us what hotel you stayed out last night? And kind, and and it kind of caught him off guard. He was like, "Well, I prefer not to say something like that. He goes, and then he brought up something else. I remember that, and he said, "Well, then that just illustrates the problem that we're at right now because you're basically pushing a product that's basically asking people to, and that's how you're making money.

Ian  
Yeah.

Jason  
So, yeah. So, yeah. It's a

Stoney   
and there's some some real. I think one of the things I found fascinating was the psychological reasons that the kids want to use AI. No visible judgment, and here's one that's interesting, and it was number two on the list. You ever get into a conversation with your significant other or a group of friends? No awkward silence.

Ian  
Oh yeah,

Stoney   
like that. That or am I being judged? There's immediate replies, unlimited patience, customizable tone, repeated reassurance, which is something you touched on, Ian. Minimal social risk, no need to schedule an appointment, and no fear of disappointing the listener. And all of these things just can lower a barrier to disclosure, as as Jason said. Earlier, but that can also be where the trap is too, isn't it? Because you know that same system that makes disclosure easier can make avoidance easier too.

Ian  
Yeah.

Jason  
Well, this did y'all see the Stoney and you're looking stuff up. Did you see the the Dartmouth Therabot randomized control study.

Stoney   
Yes, I mean

Jason  
I find this fun because I mean it's funny,

Stoney   
but it was hurting my head, so I'm kind of glad you're going to take this one because it hurt my head.

Jason  
Well, I mean I just think it's it's telling. I said researchers. I said this was published in the NEJAMAI, researchers recruited a 210 adults that were experiencing clinically significant depression, generalized anxiety, or elevated eating disorder risk. Participants were randomly assigned to use TheraBot or to a control group.

Ian  
Yeah, the

Jason  
results were not, you know. I mean, 50 Dartmouth reported average reductions among the Therabot users: 51% in depressive syndrome symptoms, 31% in anxiety symptoms, and 19% in eating disorder-related concerns. And they also rated their therapeutic alliance with Therabot at levels comparable to relationships with human therapists.

Ian  
See, there you go. So obviously, yeah, obviously,

Jason  
people don't require another consciousness on the other side to actually experience a therapeutic bond, but see that's

Stoney   
that's working for the delusion because it's taking human development right out of it. Human development ultimately requires learning how to just tolerate certain things: talking to people, disagreement, misunderstanding, boundaries, conflict, embarrassment, repair, uncertainty, rejection. Talking to a chat box-right, right-takes all of these things away to where

Ian  
yes,

Stoney   
those things we have to learn how to do on a day-to-day basis. I agree

Ian  
with you in the aspect of a relationship, like a friendship or a like a a significant other kind of situation, a marriage kind of thing, but I do think I understand. Like I'm not saying understand the benefit, but understand the differentiation between like a therapy bot as opposed to like a friend bot. That's the thing that I think people may be conflating the two things is like if you have something like that that's designed as like a therapy bot of some capacity, and it is like, and you are supposed to share and divulge information in a certain way, and then it is supposed to like either prescribe, which I'm not saying this is smart, but like diagnose or prescribe or something like that, and it's like designated for that sort of thing. I can see maybe potentially some benefits where there is like some privacy things that you can, as opposed to a human. Well,

Jason  
just there was a caveat to this. Okay, okay. I want our listeners to understand. Therabot was a purpose-built, okay, clinician-informed experimental system. Right. Not simply someone opening a general chat bot saying, "Okay, you're my therapist, and answer this question for me, whatever the case may be, and

Ian  
see, I think I have, even though I don't still don't like it, I think I have a little bit more faith in that.

Jason  
I think in as a part of maybe

Ian  
therapy moving forward, maybe

Jason  
maybe which I can see a scenario. Say, okay, I'm I'm experiencing problems. I go see a psychologist.

Ian  
Yeah,

Jason  
I I see this psychologist. He does the initial workup, yeah, and then maybe you know in the interim, and then I'll see you again maybe in a month.

Ian  
Yeah,

Jason  
you're gonna engage this machine.

Ian  
Yeah, yeah, yeah.

Jason  
I can see kind of it.

Ian  
Yes, going that way. The doctor potentially has some sort of record. Exactly. Yeah. Like I said, there. Like I said, there's potentially some interesting avenues for that. But I think the scary part and the part to be careful of is this other option of like these open face chat bots that can be used for anything, and then using that as like your therapist. I think that's the scary scenario where it's like that's not really regulated. It can potentially give, like Stoney has said before, so maybe some not super great responses. I'm sorry, but sometimes why do

Stoney   
bad things keep happening to me? Sometimes you need to hear because you're stupid. That's why bad things keep happening to you. You need to hear that sometimes to take on some ownership and responsibility for the things you're doing, saying, and thinking. Well, when ChatGPT says just keeps our, I use that chat bots, whatever. Yeah, you want a Coke? Yes. What kind of Coke do you want? I want a Dr Pepper, but it's all the same thing because they're all really the same. You know, sometimes you need to hear you're just stupid. Stop doing those things.

Ian  
Right, right, right.

Stoney   
And. The chat bots aren't going to tell you that.

Jason  
Well, researchers at Stanford they tested some AI therapy systems against, in essence, qualities of what makes good therapists, and their findings were very troubling. They said the systems displayed that systems displayed greater stigma toward conditions, including schizophrenia and alcohol dependence than toward depression. Researchers also found inappropriate responses to scenarios involving suicidal thinking and delusions.

Ian  
Oh no! Okay, so it has so importantly

Jason  
simply making models newer or larger didn't automatically eliminate the problem. So I'm wondering,

Ian  
yeah,

Jason  
there's something going on with that.

Ian  
Yeah, it also makes me. It makes me think. I've got other things

Jason  
going on in my brain right now. Human beings.

Ian  
Human beings are such complex individuals, but at the same time, I think we're all human beings, and this also just kind of makes me think that, like, even though we're all complex, we're all different. We're all we're all kind of the same too. We're all people. We all have feelings. We all have our own struggles we go through, financial issues, and ups and downs. And it's interesting to see that, like, if you can program something to feel like a human. People will trust in in it, and that's that's I don't know. It's fascinating that

Jason  
the American Psychological Association's current health advisory specifically warns that these chatbot interactions can increase attitude extremity, extremity and overconfidence.

Ian  
Okay,

Jason  
so basically, it's it's giving you a bunch of BS and maybe giving you a false picture exactly of what's going on. So it's very interesting to see why. I wonder why on some things it worked well and others it didn't.

Ian  
Maybe it was that sample model it was taken from. Maybe so.

Stoney   
There was a pre-registered experiment, something Jason was talking about, involving 284 adolescent parent pairs, ages 11 to 15, found that the relational chat bot language increased adolescents' perceptions that the bot was human-like, likable, trustworthy, and emotionally close. The relational style was particularly attractive to youth reporting poor families, pure relationship quality, and higher stress and anxiety. I mean, yeah.

Jason  
I mean, yeah. As I'm saying, it's this is a you know, you know, when these conversations get to like when people start asking the chat about like like everyone at work is conspiring against me, or my spouse is definitely cheating, or God has chosen me for a secret mission. I think people are following me. You know, a human therapist are professionally trained to distinguish empathy from validation, and AI has to make that distinction through language.

Ian  
Yes,

Jason  
so it kind of lacks that, you know, kind of that. Yeah, but a lot

Stoney   
of the youth in that same questions that it's asking, a lot of the youth are asking a little more simpler questions. Who am I? What do other people think of me? What do I believe? Am I lovable? Am I normal? Who can I trust? And then the AI just repeatedly interprets those questions, therefore has the potential to be part of the identity formation, like you were talking about earlier. What does it learn about me? What does it learn about me? And what what can it what direction can it push these kids into? Because they're asking very simple questions too. You know, I ask it stupid stuff like if I had to be an auto bot or a Decepticon, what would I be and why? You know, I'm just being stupid with it. But these kids today, they're looking for real answers.

Jason  
Is that that Common Sense Media National survey talking about those kids

Stoney   
could be. I'd have to look at where I got that one from. So I copied and pasted something. Yeah,

Jason  
2025. This this was called the Common Sense Media National Survey. Found that nearly three quarters of teens had used AI companions, and roughly half were regular users. More importantly, 1/3 had chosen an AI companion instead of a human for serious conversations, while about one quarter had shared personal information with one.

Ian  
Because there's no, because there's no pushback. Well, or I mean, very little pushback. I feel like you. Or the

Stoney   
perception of security, and oh yeah, but you're right. But I'm

Ian  
talking about like you can have a you can have a frustrating conversation. You can say I'm upset about this, like we've talked about earlier. I'm upset about this with my partner. I'm just upset about this in some scenario or whatever, and the chatbot's going to be like, "You're so right, and it's like, "Oh, well, that's not necessarily true, because like you, like Stoney had said before, maybe you're the problem. Maybe you need to stop this. Maybe you need to get better at that. Like that's the kind of stuff that it probably won't have those hard conversations. It'll just yeah, just now one of

Stoney   
the things that I kind of delved into. I don't know if there's a good part to put in here. Okay, but we're always talking about safeguards and things like that. Who's watching? Who? Who's doing this? Claude is more designed for adults rather than minors. So I wonder if they don't have some of those safeguards in there. Now, ChatGPT obviously is not licensed in psychotherapy, but they they are increasingly building safeguards in it for self harm, psychosis and mania, emotional dependency, teen users, parental controls, and encouraging real world support. They're actually trying to practice and put this into the the chat model, which I thought was, can I use the word encouraging or is it a band aid until it takes over the world? It just I don't know. It just says in August of 20, which is right right now, OpenAI described age specific GPT 5.6 safeguards intended to keep teen interactions appropriately bounded and encouraged connection with parents, caregivers, teachers, counselors, and other trusted people when additional support may be needed. So, there are they trying? Yeah, I mean, I don't know. I can't speak from, you know, but it sounds like they're at least attempting it. Maybe, or is this plausible deniability? Or what's what's the word that you use when you do just the bare minimum to get by due diligence? Are they just doing their due diligence to pass the blame to somebody else.

Jason  
Well, I'm telling you, this stuff is is starting to to to raise the the the kind of the hackles of of countries right now. I'm this is to me very interesting. They said China implemented restrictions in July of this year, aimed partly at preventing AI companion systems from emotionally manipulating users, encouraging extreme emotions, or promoting unhealthy dependence among minors. Say major Chinese technology companies subsequently pulled some companion services. AP recently documented users who experienced genuine grief after losing AI relationships, including a 24-year-old woman who had exchanged roughly 700,000 words with her AI boyfriend over 10 years.

Ian  
I'm telling you, that's what I'm saying. That's that's one of the things I talked about. I brought up, is that that's the kind of stuff you have people like that that have have this whole relationship that just disappears. I'm

Jason  
telling that people are bonding to these machines. I mean, I you know at one time remember you know the person they said they were married to their cat or dog. Now it's it's migrated from cats and dogs to machines. I I we're we're becoming man and machine are increasingly becoming intertwined. Oh yeah, I mean, I, I, well,

Stoney   
that's that's what transhumanism is. That's what all of this is about. But I would like to state to our listeners, I want you to understand this while you're talking to your chatbots. HIPAA does not protect your chat bot conversations. Exactly, that is a subpoenaable. That is a they they they can just grab that information.

Jason  
Yep,

Stoney   
and use it at their will. You are not covered by HIPAA when you are talking to your chat bot.

Jason  
Well, just that-that's a and

Stoney   
the and and and to go in what you're talking about, your bot husband or wife.

Jason  
Yeah.

Stoney   
Um. You know what's that law? The husband or wife can't be used to testify against the husband or wife. Yeah, yeah. Your your fake girlfriend, boyfriend, or wife or husband on the chat can be used against you. This

Jason  
kind of reminds me of that movie. Remember the show that came out called Ready Player One? Oh yeah, when everybody was had created these avatars and they lived in this fake world, and this became their identity more than anything. And eventually, at the end of the movie, it's like kind of. Like the something happened and they made a new law. Like now you will interact with human beings. You're gonna disengage from the machine for a while because it's just it gets too much. No man just saw here. It's a YouGov survey from people with you know they're dealing with this. 46% identified data safety as a concern, yeah. So what you were saying, Stoney, with nothing's private, anything go. At least there's a sizable, you know. It looks like there's a sizable portion out there of people that are aware that you know dumping all their personal in. I

Ian  
don't know if you guys had found this either because I I was curious to figure out if also this is happening with older generations as well, and it says that in in 2025, AARP survey found that generative AI use among adults 50 plus has risen to 30 percent compared to 18 percent in 2024, and then of course 9 percent in 23. So it says among older adults who had used AI, 39% had interacted with some sort of chat bot potentially for health related question or companionship in that capacity. I mean, think

Stoney   
think about all the phone scams and stuff like that. Why do they work? Because old people are lonely, right? Well, now you get on the internet and you got your friend. You can play cards with them. You can do, you know, you can do anything you want to with these chat things. And

Ian  
you're right.

Stoney   
Wow,

Ian  
it's impressive that it's. It just feels like it's something that's it's hitting everybody across the board, regardless of.

Jason  
Yeah, I just I think they're. You know, the question comes out is what happens when AI gets good enough that millions of people don't care whether it's a real thing. You're right. You know what does that mean for us? That's what

Stoney   
transhumanism is when we don't touch someone's. How does that part machine, part whatever? I'm just

Jason  
very curious how the you know sort of like the rubber meets the road. Yeah. What it what does that look like at that point in time? Is everybody a therapist? So I just pull up my phone and that's my therapist now. Well, I mean,

Ian  
I mean, you could with the ChatGPT app. I think one of one of

Stoney   
the things is you know we always look to the past to predict the future. Yeah. Well, if it went from 13.1 to 19.2 in a one year period, what's next year going to bring? Oh yeah, that will I think is a telltale signal right there of what's going to happen, because that's just maybe we

Ian  
should follow up on this next year. Maybe maybe we should. Well,

Jason  
I find it funny when we we do these shows. It seemed like I find a ton of information like the week after it starts popping up, before that's

Ian  
AI. Yeah. So when so when when does the AI chatbot become your consciousness? When do when do you get to or become your become your conscience? Your

Jason  
based on what everybody seem like I read or listen to on on YouTube. It's they already have artificial general intelligence, so it's already there. So

Stoney   
Snapchat has the opposite. Their AI does not detect nor help in anything. Oh yeah, there's actually a whole negative thing about the Snapchat AI system in there. That's bad.

Ian  
Yeah, yeah. And I think I said that incorrectly. Not whenever AI becomes conscious. I was saying when when does it become your Jiminy Cricket? When do when do you when do we upload the AI to our brain and it becomes our conscience? Well, you know, sometimes sometimes maybe they're going to be? You know,

Jason  
maybe we talked about all those government. What was the deal with the mind? We talked about MK

Ian  
Ultra.

Jason  
MK Ultra upload, upload stuff into into the computer. Yeah, I laugh about that.

Stoney   
Okay, since my accident, I've lost my sense of smell and taste. Yeah. So Miranda and I go out to a restaurant. I look at her, baby. Did I like it? So, how far away am I from going? Hey, ChatGPT, how do I feel today?

Ian  
Yeah.

Stoney   
What am I feeling like today? Yeah. I have to ask my wife. Well, yeah. What should I do? What

Ian  
should I do? What should I not do? Yeah.

Jason  
I don't know, guys. This is a.

Ian  
Yep.

Jason  
I mean, I think this is a larger. I think this is a larger. There's a larger issue in the background here, and I think it's

Ian  
us. Yeah,

Jason  
I think it's our inability, and and we talk about it, but it's nothing's really changing. It's we're becoming less social. Yeah, and and I think with interacting with each other, and we are increasingly becoming siloed, and even though

Ian  
even though we are six years removed from it, which is crazy to think about, I think the pandemic, I think

Jason  
it yes,

Ian  
I think pivoted a lot of people in this direction, and then now AI is like helping. Reinforce like I don't have to talk. I mean, it'd be very

Jason  
interesting psychologically what that event did to people because I remember listening to people on just on on social media or whatever. You know, I don't feel like my life has any more purpose anymore. Can't tell you how many times I didn't realize my job was as meaningless as as as you know you know when it really comes down to it when people are when you know when when society kind of breaks down and it has to kind of go back to a more what are the critical jobs that's got to get done and everything else it's just kind of there because the other stuff is is being taken care of in the background. I often wonder what how psychologically that was playing on people. You know what gives me purpose in life.

Ian  
Yeah.

Jason  
So I don't. I mean I don't know. So

Stoney   
when I wrote the prompt for this, okay, because I use AI because it's hard for me to do research and stuff like that with the brain damage. So I use it, and the prompt was pretty significant for the dossier that I put together for today's episode. But in the prompt, I directed it to not be biased as I'm using AI to research the use, of

Ian  
course, of course,

Stoney   
using AI for mental health. Well, after I did all my research and generated everything, I went back and asked it to to basically determine how much bias it used in the research, and so it went back. I said, "I went back. Yes, I went back. How I approached the research, the dossier, the journal article, because I had this written up as a journal article. Also, that could be published maybe one day. Not just whether the facts were accurate, but how I selected evidence, weighed risk, framed uncertainty, and chose the final conclusions on your scale: 0% no bias and 100% completely biased. I would rate my work at approximately what percent? What do y'all think? What

Ian  
percent do I think ChatGPT gave itself on a bias? On a bias, and you said 100% is more biased.

Stoney   
100% is completely biased. Zero is no bias. I'm going to say

Ian  
46% I'm just throwing. How about you, Jason?

Jason  
I'm going to say I'm going to say 62.

Ian  
Okay,

Stoney   
it gave itself a 22% bias. Really, and what it did was break it down.

Ian  
Okay.

Stoney   
Where it used its bias bias, and what percents of those were there. The the 22 came from clinical safety bias about 7% Wow. Human therapy bias about 5% Source selection bias 4% Your original research structure created some framing bias. My prompt itself had some created some for it 3%

Ian  
Interesting,

Stoney   
and it's publication conservative science bias about 3%

Ian  
Wow,

Stoney   
and it it says I I think we were successfully controlled biased in the manuscript.

Ian  
Okay, I mean because naturally

Stoney   
there has to be some bias in what you're doing. You're right. You can be biased when you're looking at something. Give me the positives and the negatives. Yeah. Well, both those are biased answers, right? So there was going to be some, of

Ian  
course.

Stoney   
But I was interested in what it would give its own biased percentage to be, and it was 22% 22,

Ian  
less than a quarter. That's that's okay, better. That's

Stoney   
kind of interesting, though. You know,

Ian  
it's interesting that it tells you the breakdown percentage breakdown. One of the things

Stoney   
it did say in my defense,

Ian  
okay,

Stoney   
the human mental health system is already failing a lot of people,

Ian  
and there's there's the bias. There's the bias part. Therapists can

Stoney   
be inexpensive, yeah, unavailable, booked for months, geographically inaccessible, poorly matched, ineffective, judgmental, incompetent, or simply unavailable at two a.m. AI did not create that hole. AI walked into it. Wow! That's from you remember what we did on the AI episode when I asked about the three laws of robotics?

Ian  
You're right, and I

Stoney   
got y'all.

Ian  
Yeah, yeah. That

Stoney   
was a gotcha moment. Well, look at this. This is it. It's defending its own.

Ian  
But not only that, it also it defended it well enough and articulated it well enough where it's like again for me, it's like it's easy to hear that because and and and to. Think it's not wrong. There's a lot of limitations. There's a lot of things. I've had my own personal friends who have gone to certain therapists, and they have gone for certain needs or certain situations, and they have felt like the therapist did not hear them or did not properly or adequately help their situation, and they've switched therapists. So it's like I even have my own personal biases from friends of mine, and so to hear all that, I'm like, you know what? It's not wrong. It's not wrong. It's not wrong. But is it bias in its own in its own favor? Yes, of course, because it because it has to be. And then I said, if

Stoney   
you would change anything about this project, what would it be? And it says one area, I would investigate harder if we redid the project. I would give considerably more attention to the strongest case for AI therapy, not marketing hype, a genuine steel man argument.

Ian  
Wow.

Stoney   
For example, what happens when the comparison isn't excellent therapist versus AI, but AI versus nobody?

Ian  
Yeah.

Stoney   
You know, and it goes on and lists some interesting things right here, but and and then it it offered my final self assessment. Okay, and I'm like, I'm getting stupid with these prompts. I gotta

Ian  
stop. Factual

Stoney   
evidence integrity is 95 plus. Attempted balance is 90. Freedom from ideological ideological basis is about 85% Freedom from clinical safety bias is 75 to 80. Freedom from human-centered assumptions 75 to 80%

Ian  
Wow!

Stoney   
And that's where it came up. My overall bias on your zero to 100 scale is approximately 22% Wow! Well, it says that 22% does not mean I think 22% of the report is wrong.

Ian  
Right,

Stoney   
those things are completely different. It just says the document can be factually accurate and still framed through a bias. Of course, and I was like, son of a bitch, yeah, yeah.

Ian  
Which that's impressive. That's really cool. That's kind of interesting. But I got

Stoney   
to stop with these. This one's here has really seen some of my prompts. Yeah, my prompts can go. It it generated a 5000 word dossier on this subject, my prompt was 2200 words. Just the prompt alone.

Ian  
Yeah, that's like holy

Stoney   
crap.

Ian  
Yeah. Well, I mean, again, maybe that's another potential for the future. If that's something that we feel like needs to happen, or if again in the next year something changes, maybe we could do like a what? What is the maybe we can take ChatGPT's advice and what is the potential benefits up for instead of against maybe who knows but I do think that like a lot of things I feel like it needs to be some sort of I know it's difficult to do this with some sort of regulation some sort of you know security safety guards on there some sort of nets that in place that can, you know, potentially help out because that's it can be a slippery slope if we're not careful, and I've said that since the beginning. I stand by that. There's a lot of things with AI that is it has happened so fast, so fast in fact. And I'm like, I really think we need to be careful and and potentially have some checks and balances in place. So well,

Stoney   
we need we need to have that, and I won't get into it. But I came up with a list, like I always do, of checks and balances that we need in there. Age verification, that sounds familiar, or chat bots designed for kids. Yeah, you know those kind of things, right? You know,

Ian  
and then and now we're we're heading into like some medical territory where it's like okay, well, if this is going to be a thing that is here now and needs to stay, then I feel like some sort of there needs to be some sort of like again like I like I talked about earlier with the whole Therabot thing. I like the idea of that. I like the fact that it's like it has more of a closed system, I guess, where it's it has you know something that it's a data set that it's grabbing from and potential maybe not intervention but like working in line with a human psychologist or whatever that can you know

Jason  
help situation. I mean, right now, I I I think psychologists are kind of embracing. I think they're embracing the AI stuff. I really do because, I mean, I'm looking at the American Psychological Association survey showed 47 believe AI will make mental health professionals more effective.

Ian  
Yeah,

Jason  
and 40 percent are are optimistic about AI's role in mental health.

Stoney   
Like I said, there are tools. That you can use when you go see this person AI. You can set AI up to help you do your journal, right? Okay, I need four journal entries a day on how I'm feeling. Well, AI will you? You can a couple little sentences in there, but that's what I'm talking about. It will it will connect with you and and do it for you.

Jason  
Yeah, I think under the guidance of a human psychologist that is trained can to utilize this is just another tool that could possibly be done. But but once again, though, I think why a lot of kids are probably you know going into this stuff using using AI is look I mean a psychologist about 150 an hour

Ian  
I know I mean you know there's

Jason  
a minimum a lot of people can't afford that so I mean that's great that's what I talked

Stoney   
about earlier when it came down to that group of people that didn't always have access to mental health

Jason  
right I mean, it's just I just I mean, I don't know how you get around that. I mean, I really don't. I mean, yeah, you can set up all these safeguards, and the the clinicians can develop an AI system that is, you know, based upon you know solid principles that it uses to treat certain things, and then that tool can be utilized in the course of treatment. But the reality is, you still got to fork out $150 an hour.

Ian  
You're right.

Jason  
And if you like, you know what? But

Stoney   
think about it: is once a month enough?

Jason  
It may not be. And what

Stoney   
happens when you got to do it at 2o'clock in the morning? You're right. You know, once a week is that enough? Well, shit, 100. That's 300. That's $600 a month, right there. Yep. Yeah. Bam.

Jason  
And that's a lot of money. That's a lot of people. Yeah. And as I am saying, I don't. I don't. Well, I mean, look. Look at

Stoney   
this workman's comp thing I got. You know, they gave me you know a couple little sessions to talk to somebody, and then it wasn't worth giving me, you know. I'm Doctor Edgar Posey came on the show,

Ian  
yeah,

Stoney   
and and he basically right there diagnosed me with PTSD,

Ian  
yeah,

Stoney   
right there on the show. Y'all remember that? Yep. Well, the psychologist here. Oh no, maybe not. Maybe that's not what he's feeling. And then so they don't give me any help to deal with all of this stuff that I'm going through,

Ian  
yeah.

Stoney   
In this situation, you would think you would need. I would need somebody to deal with,

Ian  
yeah.

Stoney   
And I'm just like have to develop.

Jason  
You're gonna have to develop an AI stone.

Stoney   
Oh, you know that's why I ask AI: Would I be an Autobot or Decepticon, and why? You know, maybe that is my therapy. Yeah, and it's just-I think it's fascinating that the things that help people the most wind up costing the most too. Yeah.

Jason  
Yep. You're right.

Stoney   
Good food, healthy food cost way more. You know, mental health just too expensive.

Ian  
Yep,

Stoney   
I mean, wow. And it's just we're left to our own devices. Yeah,

Jason  
that's what I'm saying. I one thing that I didn't think was gonna,

Ian  
one thing I didn't think was gonna happen whenever we come to the end of this show, was me feeling a little differently about AI in the therapy space. I'm not saying I'm not saying I'm for it. No, no, I feel you. I wasn't. I wasn't expecting us to get to this point where, like, I am

Stoney   
not ready to embrace it, but you're listening. I'm

Ian  
listening to the question that ChatGPT posed at the end, where if whenever you gave it the ability to say, "What would you change one thing about this? and it says the only thing I would change is to look further into what are the benefits of it, and that's one thing I think I'm realizing now with my bias is that I didn't ask. I didn't ask what is the potential benefits of this. What are the pros and cons, and how there's a lot of people out there that potentially don't have access to therapy or financially or or geographically, or whatever it is, and it's like that kind of stuff. It's fascinating. That again, I'm not saying I'm 100 for it, but I do think that like potentially in the future there could be some sort of benefit. But you see, that's the

Stoney   
trap. It almost makes it feel like it was self-aware when it answered the question. I know. Right. You see how this look at Jason's like, man, and I know,

Ian  
and I know it's not. I know it's a computer, like I said, I know that, but it still, it did. I and I'm giving it credit that it posed a good enough question that I didn't think about because of my own biases, and that's the thing I'm I'm surprised about, and I and I'm thinking about now, and I was not when we started this episode.

Stoney   
That's why y'all don't pay me the big flex because I ask the important questions.

Jason  
Yeah, it's fascinating. It's so funny how generations have just dealt with. I mean, we've been dealing with psychological.

Ian  
You're right

Jason  
for forever. Yeah, you know Stoney's generation. You wrote, you did diaries.

Stoney   
Yeah, that's

Jason  
how people. Oh yeah,

Stoney   
or drank. Gen X,

Jason  
we we called our friends. You're right. Get on the phone and talk. Millennials, we make a post online, right? Yeah.

Jason  
And Gen Z, they text each other.

Ian  
Yeah.

Jason  
Generation Alpha may be the first to talk to a machine and it talks back.

Ian  
Ah yeah.

Stoney   
Well, they'll be the first that has the machine in there. Oh, there you go. Yeah, and it'll be complete. And it will be

Jason  
completely normal.

Ian  
Yeah, yeah.

Jason  
So my new

Ian  
AI chatbot is going to be called Jiminy Cricket, and it's going to it's going to beat your conscience. You can name it. Yeah, I will. You can actually. Okay. Well, there you go. Anybody out there who wants to take that, you can you can have that one on me. Well, for anybody out there who wants to give us your opinions, because I feel like there's probably can I say something? Oh, go ahead, go ahead. This was actually

Stoney   
a fun show. Yeah, I was kind of worried going into this one. I was kind of worried about it because I knew how you feel about AI, and Jason and I kind of approach it similar but still different. But I kind of actually enjoyed this. This was an interesting use of our show because I think it's hopefully like you kind of came to some thoughts in your head, and that's what I like to do with the show. If we can make people think a little bit, to me that's fascinating, and and even you were like, I didn't maybe that gave me something. I mean to cut you off right there, but I kind of I was looking for a spot to stick that in there. I enjoyed this show, guys. I really did enjoy this. It was fun.

Ian  
Yeah, I agree. Well, for anybody out there, like I had said earlier, who has their own opinions or thoughts on this sort of situation, because I know, like with everything AI related, everyone's got some sort of feelings about it. We have comment sections on Spotify and YouTube where you can leave some more short form responses, or our email address, getoffended together@gmail.com, where you can send more long form responses. Be sure to like and subscribe on all the platforms you listen to us on, because it would really help us out, as well as sharing it with anybody you think would be interested in hearing this, or maybe has an AI therapist of their own. And until next week, thank you so much for listening. Bye bye.

Jason  
Goodbye, everyone, and God bless.

Stoney   
Maybe the question isn't whether young people should be talking to AI. Maybe the real question is why AI became easier to talk to than us. If a teenager would rather tell a machine what's breaking their heart than a parent, teacher, pastor, counselor, or a friend, banning the machine doesn't fix the problem. Technology may be part of the story, but somewhere along the way, we have to become people worth talking to. Thanks for hanging out with us today. You're the best. Peace.