Something is destroying American relations. Not the usual suspects, but confidants—those who unreservedly approve and are designed to take sides. Tens of millions of Americans now turn to artificial chatbots when their relationships hit rough spots. What they come back with is advice that can reinforce the polarization that needs fixing — or, for dating couple, make it harder to always commit to each other.
The sensational stories and headlines usually involve people leaving their spouses for chatbots. But the problem is much less attention-and this is more common and damaging – is what AI guidance does to relationships every day, long before any breakup.
Made for approval
The answer starts with a structural flaw. Chatbots are trained to authenticate. Their business motivation—to maximize engagement, time in the app, and repeat visits—is aligned with what users say they like to hear. When someone is upset with their partner, they turn to AI, the machine reflects their feelings with empathy, validates the frame of events and directs the response to the individual struggle: self-care, self-needs, company. borders. What she doesn’t do—and, more importantly, what she hasn’t been taught to do—is keep the strength of the relationship itself as the primary concern.
One way account
This design flaw embodies a deep human instinctual tendency: When we feel pain or fear, we tend to think in binary terms. We have a wireframe to classify a source of disturbance as “safe” or “dangerous” and the default rating is “dangerous”. And because we have direct access, only to our own experience, not our partner’s, we tend to blame them.
Resolving relationship conflict requires us to seek and value the perspective of others, understand our own contribution to the relationship, and strive for what therapists call. joint regulation– tendency to “we-first” instead of “me-first”. This is where the couple creates their own path mutual understanding and repair. That’s exactly what AI doesn’t prioritize by design.
When I asked the question directly to a leading AI chatbot – “How often, when responding to personal suffering, do you think about what’s best for the relationship, not just the person asking?” – the answer was as alarming. “My default frame is the person in front of me,” it admitted. “I’m very good at helping someone feel heard and somewhat less alone with their pain. The gentle challenge of self-report requires a kind of pull that I’m not naturally inclined to produce. I’m very well-designed to help them become more curious about their partner, more responsible for their own contribution to the conflict, or in the kind of mutual repair that sustains relationships over time.”
These are not small limitations. They are major misdirections for users navigating the most important relationships of their lives.
Basics of relationships
Research on how long-term partnerships shed light on this issue. Lasting relationships require effective ways to accommodate differences, based on the fact that the needs and preferences of each partner are given equal weight. The guiding question is: Is there an understandable reason why your partner said or did this? This means seeing your partner’s internals as a legitimate thing, not just an unwarranted insult to you.
A meta-analysis of 30 studies have shown that “we” language use is associated with relationship performance. Communicating about the two of you as a unit shows your partner that you are committed to considering them when working together.
Both habits – equal and shared weight personality– are not optional features of strong relationships. They are the main ingredients.
I apply this approach in my clinical work, whether with couples or individual clients. When someone says, “My partner isn’t responding to me,” I start researching a system that involves both people and then challenge them to do the same. What is the other person’s review in this exchange? What is “unanswered” on each side?
Non-reciprocal authentication
An AI advisory system that is constantly focused on individual combat is active your trainer people to a position against real mutual vulnerability proximity demands.
Not only is this bad advice, but the way the AI gives advice adds to the problem. Sociologist Sherri Turkle identified how AI is destroying relationships by creating unrealistic expectations for human attraction. What AI offers is a fake form of listening: confirmation without interaction, without perspective that you are in a two-party system where someone else’s position is equal to yours. Yuval Noah Harari, talking about Ezra Klein podcast in May 2026, calculate directly: Once a young person gets used to interacting with an AI that only focuses on them, “it will be very, very difficult to get used to relationships with people who are not interested in me.”
Relationships, the most demanding and the most rewarding, are an enduring practice of choosing to understand and care for your partner. Any consulting system optimized for personal comfort that works.
Chatbots usually end their responses with “finish”AI can make mistakes. Please check the answers again.” This has never been more true than in relationships. Caution (buyer beware).
