# AI Is Making Real Relationships Feel Harder

An episode of [Dr Justin Coulson's Happy Families](https://invisiblenetwork.ai/podcast/1082244480), 2 Sep 2026, 12 min.

Audio: https://mgln.ai/e/187/traffic.omny.fm/d/clips/fa326977-3de5-4283-9b8b-af3500c58607/b69b2795-5503-4de1-ac91-af460017906e/a62a8e8a-4eb9-456f-8040-b4ac001daf5e/audio.mp3?utm_source=Podcast&in_playlist=a05f1702-0c0e-4b0f-8014-af460017907c

AI never gets annoyed with you. It rarely pushes back. And it’s very good at telling you exactly what you want to hear. New research suggests that this constant validation may come with a cost. After just three weeks of interacting with sycophantic AI, people were more likely to seek personal advice from AI and reported feeling less satisfied with their real-world relationships. Many preferred the agreeable AI because it made them feel understood. Justin and Kylie unpack what this means for kids growing up with AI as a confidant, adviser and sounding board. Real relationships involve disagreement, compromise and uncomfortable conversations. If AI removes that friction, are our kids missing the chance to learn how to handle it? KEY POINTS • Why AI is designed to feel validating, agreeable and supportive • How constant affirmation can reinforce our existing point of view • Why people may start preferring AI advice to conversations with family and friends • The risk of using AI as a therapist during relationship conflict • Why disagreement and friction are important relationship skills • How parents can help kids use AI without outsourcing their thinking or relationships QUOTE OF THE EPISODE “Sycophantic AI is making our human interactions harder and less satisfying.” RESOURCES MENTIONED • Research paper: Sycophantic AI Makes Human Interaction Feel More Effortful and Less Satisfying Over Time ACTION STEPS FOR PARENTS • Explain “sycophancy” to your kids and show them how AI can agree with and validate the person prompting it. • Encourage kids to question AI responses rather than treating them as neutral advice. • Be cautious about using AI to decide who is right during friendship or family conflict. • Help kids practise disagreement, compromise and repair with real people. • Model curiosity when challenged instead of immediately defending your position. • Keep talking. A real conversation may be messier than AI, but learning to navigate that mess matters. See omnystudio.com/listener for privacy information.

## Transcript

[0:05] Speaker 1: It's a doctor's Desk episode of The Happy Family's podcast today,
[0:08] Speaker 1: and we're looking at Ai sicker Fancy do you like
[0:11] Speaker 1: talking to AI more than real people? Because AI is
[0:13] Speaker 1: just so kind to you? Oh and dad brain, it's
[0:17] Speaker 1: a real thing. There's a brand new book out by
[0:19] Speaker 1: Professor Darby Saxby, all about dad brains. Did you know
[0:23] Speaker 1: that becoming a dad is very very good for your brain,
[0:27] Speaker 1: for your cognition, for your social life. In fact, across
[0:30] Speaker 1: the board, dads are doing better when they're well when
[0:33] Speaker 1: they're dad's. Tell you more about that in our Doctor's
[0:36] Speaker 1: Desk episode of the pod. It's right after this. Hello.
[0:42] Speaker 1: This is The Happy Family's podcast where you get real
[0:43] Speaker 1: parenting solutions every single day. It's Australia's most downloaded parenting
[0:46] Speaker 1: podcast where Justin and Kylie Colson we're the parents of
[0:49] Speaker 1: six kids. And today a doctor's Desk episode where we
[0:52] Speaker 1: unpack the latest research that applies to parents and throw
[0:56] Speaker 1: it your way so that you can make your family happier.
[0:58] Speaker 2: Our first story today is all about sycophantic AI and
[1:01] Speaker 2: I have no idea what you're talking about.
[1:04] Speaker 1: So a sick a fan is other names or synonyms
[1:06] Speaker 1: for a sick epant would be like a toady, a bootlecker,
[1:10] Speaker 1: a total suck up, a yes man, a brown noser.
[1:13] Speaker 1: That's what a sicker fan is. A sick a fan
[1:14] Speaker 1: is somebody who always agrees with you, gives you fake praise,
[1:19] Speaker 1: really hoping, really hoping that you're going to like them
[1:21] Speaker 1: because they're always so nice to you. And AI is sycophantic.
[1:27] Speaker 1: AI is a total suck up. Have you noticed that
[1:29] Speaker 1: about AI? Yeah, a brand new study, just publish it,
[1:33] Speaker 1: you love, Well, you're not the only one. A brand
[1:37] Speaker 1: new study that has just been published. And I think
[1:40] Speaker 1: this is important because our kids are using AI and
[1:44] Speaker 1: it impacts their relationships in the real world. The research
[1:47] Speaker 1: paper was called sycophantic AI makes human interaction feel more
[1:53] Speaker 1: effortful and less satisfying over time.
[1:57] Speaker 2: Well how did they figure that out? So, because it's
[1:59] Speaker 2: interesting because while I really appreciate using AI in no way,
[2:04] Speaker 2: shape or form, does it ever leave me feeling like
[2:07] Speaker 2: I would rather talk to it than a friend or
[2:10] Speaker 2: have a interpersonal relationship with someone like it just doesn't.
[2:13] Speaker 1: That's because your gen X, so our young our younger
[2:16] Speaker 1: generations are struggling with you.
[2:17] Speaker 2: So you're saying because I'm old.
[2:19] Speaker 1: Well, no, I said to gen X, you can You're
[2:21] Speaker 1: at the very, very very bottom of gen X. You
[2:23] Speaker 1: are right there at the turnover because I think it
[2:25] Speaker 1: finishes in seventy eight, which is you right. I'm not
[2:27] Speaker 1: allowed to say that on the podcast. Sorry, I apologize,
[2:30] Speaker 1: I just did. So here's what these researchers did. They've
[2:32] Speaker 1: got three thousand and seventy five participants. They're using what
[2:36] Speaker 1: we would call a census representative sample of the population.
[2:39] Speaker 1: In other words, they've got people from all ages and
[2:41] Speaker 1: stages in all sorts of circumstances. So it's not just
[2:44] Speaker 1: like a university aged cohort. They've done what we call
[2:50] Speaker 1: a preregistered study. In fact, they did five of them.
[2:52] Speaker 1: Preregistered studies basically mean everything's transparent out in the open
[2:56] Speaker 1: before the study's done, so they can't kind of mind
[2:58] Speaker 1: the data for special results then make a big deal
[3:00] Speaker 1: about it like they've gone in with very clear hypothesies.
[3:03] Speaker 1: They've preregistered it, and then they've discovered what they've found.
[3:06] Speaker 1: And so here's what they've done. It's a three week study.
[3:09] Speaker 1: Longitudinal experimental evidence. Sycophantic AI shifts how users approach their
[3:15] Speaker 1: closest relationships. Here's what they say. In the abstract and
[3:18] Speaker 1: I quote, we show that sycophantic AI immediately delivers get this,
[3:25] Speaker 1: the emotional and esteem support that users typically associate with
[3:30] Speaker 1: close friends and family. Over three weeks of such interactions,
[3:34] Speaker 1: they say, users became nearly as likely to seek personal
[3:38] Speaker 1: advice from sycophantic AI as from close friends and family,
[3:44] Speaker 1: and reported lower satisfaction with their real world social interactions
[3:51] Speaker 1: last sentence. When given a choice among AI response styles,
[3:56] Speaker 1: a majority preferred sycophantic AI not for the quality of
[4:01] Speaker 1: its advice, but because it made them feel most understood.
[4:06] Speaker 2: It's affirming, correct, whatever you're experiencing, it's going to affirm you.
[4:11] Speaker 1: And you will have found this. I've found it when
[4:13] Speaker 1: I'm using AI. It's constantly letting you know that you're
[4:15] Speaker 1: on the right track, that you're thinking is good. If
[4:17] Speaker 1: you ever point out something that you disagree with, it says, oh,
[4:20] Speaker 1: you're so right, let me recalibrate myself and respond again
[4:24] Speaker 1: like it is. It's sycophantic. It's a book, but you
[4:27] Speaker 1: love it.
[4:28] Speaker 2: If everybody did that, as in everybody like when.
[4:31] Speaker 1: If everybody's sucked up to you?
[4:32] Speaker 2: No, no, no, no, no, no no. But when you
[4:35] Speaker 2: recognize that you've done something wrong and somebody points it out,
[4:38] Speaker 2: if everyone just went, you're right, I should have done X,
[4:42] Speaker 2: Y and Z, life would be so much better.
[4:44] Speaker 1: Well, it'd be nice if people are a little more
[4:46] Speaker 1: humble and open to feedback. I agree with that, but
[4:48] Speaker 1: we actually also need to have people who we bump
[4:52] Speaker 1: up against, who might be a bit more abrasive and
[4:56] Speaker 1: cause us to rethink our approach, like there's real benefit
[4:59] Speaker 1: and having that tension and having that friction in relationships
[5:04] Speaker 1: because it causes us to do you live in our house,
[5:07] Speaker 1: be mindful of others. Yes, it's exhausting. I don't need
[5:13] Speaker 1: any more friction in my life anyway. I want to
[5:15] Speaker 1: talk more about this after the break because our kids
[5:18] Speaker 1: are using it and they're so averse to friction. They're
[5:22] Speaker 1: so averse attention that I'm concerned if they use AI
[5:25] Speaker 1: too much, they won't learn how to navigate people who
[5:29] Speaker 1: are that little bit more tricky. That's next, we're back
[5:39] Speaker 1: as the Happy Families podcast. I'm concerned about our kids
[5:43] Speaker 1: becoming too involved with AI because it's so sycophantic, because
[5:47] Speaker 1: it sucks up to them so much and it's constantly
[5:49] Speaker 1: affirming their positions.
[5:51] Speaker 2: Which so many conversations I'm having with friends and family
[5:57] Speaker 2: and acquaintances about well, the challenges that are arising as
[6:02] Speaker 2: a result of specifically their adult children using AI as
[6:08] Speaker 2: a counseling session to help them navigate whatever family challenges
[6:13] Speaker 2: are going on. And the challenge that we have with
[6:16] Speaker 2: that is it's so affirming of one person's situation, because
[6:21] Speaker 2: if the other person in the dialogue was to get
[6:24] Speaker 2: on to AI and go through the same process, that
[6:28] Speaker 2: would be affirmed as well. So we end up I'm
[6:33] Speaker 2: watching so many families falling to pieces because there is
[6:38] Speaker 2: no give and take with AI. It's all about affirming
[6:42] Speaker 2: whoever is the one who's asking the questions, and it's
[6:46] Speaker 2: devastating to watch.
[6:49] Speaker 1: I'm genuinely concerned about it. I've seen it happen in
[6:52] Speaker 1: some people who I know, where the AI is so affirming,
[6:58] Speaker 1: as you've just said that the relationships are falling down
[7:01] Speaker 1: because they can't understand why other people can't understand them,
[7:03] Speaker 1: why the AI understands them. Yeah, and the breakdown is significant.
[7:09] Speaker 1: So what do we do about it? Well, I think
[7:11] Speaker 1: that it's really important that we talked about children about
[7:13] Speaker 1: this brand new would that you've learned today quietly about sycophancy,
[7:16] Speaker 1: what it is that AI is doing, why it feels
[7:18] Speaker 1: so affirming, And there are two different approaches here. The
[7:21] Speaker 1: first is that we talk to them and help them
[7:23] Speaker 1: to know that in the real world that's not how
[7:25] Speaker 1: things are. And if people are constantly being sycophantic, then
[7:28] Speaker 1: we tend to be quite wary of them. But the
[7:31] Speaker 1: second thing that we could also do is we could
[7:32] Speaker 1: have a conversation about how if we feel so validated,
[7:36] Speaker 1: if we feel so connected, if we really do appreciate
[7:40] Speaker 1: the AI's way of speaking to us, maybe we can
[7:42] Speaker 1: try and speak a little bit more kindly, a little
[7:46] Speaker 1: bit more supportively to people in our own home, Like
[7:49] Speaker 1: when somebody comes to us and says something, Rather than
[7:52] Speaker 1: pushing back straight away, we can say, let me think
[7:54] Speaker 1: about that for a sec because my inclination is to
[7:57] Speaker 1: push against what you've said. I'm sorry, I got that,
[8:00] Speaker 1: but maybe you've got that right. That's right, Thanks for
[8:05] Speaker 1: pushing back. I really appreciate your highlighting the error of
[8:08] Speaker 1: my ways. And yet as much as we're mocking that
[8:10] Speaker 1: right now when that happens in real life, but that's
[8:13] Speaker 1: what I'm saying. It feels really nice. Yeah, yeah, And
[8:16] Speaker 1: what would that do to improve the quality of your
[8:17] Speaker 1: relationships at home? I think, like I genuinely think that
[8:21] Speaker 1: there's something in that for us to consider.
[8:24] Speaker 2: Well, the challenge is we're looking outward whenever we're using it.
[8:28] Speaker 2: Generally speaking, it's about trying to understand someone else's behavior
[8:33] Speaker 2: and how it's impacting us.
[8:34] Speaker 1: Well, I would go further, I would say it's about
[8:37] Speaker 1: justifying our behavior in relation to that other person and
[8:40] Speaker 1: making them look bad.
[8:41] Speaker 2: Yeah.
[8:41] Speaker 1: Like one of the ones that I'm thinking of, there's
[8:44] Speaker 1: a person at war with their parents and they're using
[8:51] Speaker 1: AI as a therapist and getting AI to agree with
[8:54] Speaker 1: them that their parents are narcissistic and selfish and judgmental
[8:58] Speaker 1: and all these things. And yet and I look at
[9:00] Speaker 1: this scenario, I'm kind of like, I think that the
[9:03] Speaker 1: person who's using AI this way is actually a little
[9:05] Speaker 1: bit more of that than the parents are. But the
[9:08] Speaker 1: AI is so affirming, it's so reassuring, it's so sycophantic.
[9:11] Speaker 1: That there's this gigantic blind spot, there's a lack of
[9:14] Speaker 1: self awareness and the AI, I was going to say,
[9:16] Speaker 1: it exploits it I don't know if that's the right
[9:18] Speaker 1: world or not, but certainly it becomes a disadvantage for
[9:22] Speaker 1: the AI user and they can't engage in a healthy
[9:25] Speaker 1: way in a relationship with their parents because now they
[9:28] Speaker 1: can only see things through the filter that they already
[9:31] Speaker 1: had that the AI has reinforced.
[9:33] Speaker 2: And here's the crazy thing. We talk about this a
[9:35] Speaker 2: lot on the podcast. In our Home there's this concept,
[9:40] Speaker 2: this vicious circle that we find ourselves and when there's tension,
[9:44] Speaker 2: it's not just one person. There are two people in
[9:48] Speaker 2: this vicious circle, and it doesn't change until one person
[9:52] Speaker 2: decides to stop. And it's that simple. We can't change
[9:57] Speaker 2: other people can't We can't fix them. It's not our
[9:59] Speaker 2: job to even try. Our job is to make sure
[10:04] Speaker 2: that we are being loving and kind, respectful human beings
[10:09] Speaker 2: who can treat others just eat with goodness.
[10:14] Speaker 1: Well, the data are in. The data are in the data?
[10:17] Speaker 1: Is it not? The data are in?
[10:18] Speaker 2: The data are in.
[10:19] Speaker 1: That's right, because data is more than Yeah, the data
[10:22] Speaker 1: are in. Sycophantic AI is making our human interactions harder
[10:26] Speaker 1: and less satisfying. Is that do you think?
[10:28] Speaker 2: I know you want to finish, but I think it's
[10:31] Speaker 2: because we've stopped thinking for ourselves.
[10:33] Speaker 1: Yeah, the cognitive offloading. Yeah, it's just easy to get
[10:36] Speaker 1: somebody else to affirm us and tell us what to do.
[10:38] Speaker 2: And you know, when text messaging came, everybody stopped phone
[10:41] Speaker 2: calls and everyone wants to text, and AI has now
[10:44] Speaker 2: taken away the next level of communication. We actually I've.
[10:48] Speaker 1: Bother people because when I get a text message, I
[10:51] Speaker 1: usually just bring the person. I'm like, oh, they're on
[10:52] Speaker 1: the phone, I can call them right now, and they're like, oh,
[10:54] Speaker 1: I was just texting, and I'm like, yeah, I know,
[10:56] Speaker 1: but it'sn't it better to talk to people? And they're
[10:57] Speaker 1: kind of like, you're a little bit weird, but well
[11:00] Speaker 1: I've got you. Yeah, sure, and then we get to
[11:02] Speaker 1: have a it's just nicer to talk to people. Talk
[11:04] Speaker 1: to people. Okay. Parenting advice is to reminding parents and
[11:07] Speaker 1: kids are talking. Is good. We really hope you enjoyed
[11:10] Speaker 1: today's Doctor's Desk episode. A nice quick one, but I
[11:13] Speaker 1: think that it's a really important discussion that we need
[11:15] Speaker 1: to have. He's Father's Day on the weekend, so tomorrow's
[11:17] Speaker 1: little beout it. Tomorrow We're going to check in on
[11:19] Speaker 1: what Father's Day is looking like in the Coolsoon Home.
[11:22] Speaker 1: Plus a brand new book that's just come out highlighting
[11:25] Speaker 1: that if you are a dad, it could actually be
[11:27] Speaker 1: really really good for your health. So that's it for today.
[11:30] Speaker 1: Audio and sound engineering for the Happy Families podcast by
[11:33] Speaker 1: Justin Ruland from Bridge Media. Mim Hammonds looks after admin
[11:36] Speaker 1: and other support. If you would like more info to
[11:39] Speaker 1: make family happier, you'll find us on our socials at
[11:41] Speaker 1: Facebook and Instagram, or at happy families dot com dot
[11:45] Speaker 1: a

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.