Something is hurting American relations. Not the usual suspects, but trusted confidants—those who tacitly confirm and are designed to take sides. Tens of millions of Americans turn to AI chatbots when their relationships hit a rough patch. What is repulsive is advice that can reinforce the polarization that needs correction get to know couples find it difficult to bond with each other.
The buzzy, headline-grabbing stories usually involve people leaving their spouses for chatbots. But the problem has received little attention attention-and this is more common and harmful – this is what AI guidance does to relationships every day, long before any breakup.
Created for approval
The answer begins with a systemic flaw. Chatbots are trained to validate. Their commercial incentives—increasing engagement, app usage, and return visits—are consistent with telling users what they like to hear. When someone who is upset with their partner turns to AI, the machine empathically reflects their feelings, confirms the events in their structure and focuses the response on how the person behaves: self-care, own needs, firm. boundaries. What he doesn’t do—and most importantly, what he’s not taught—is to hold the strength of the relationship as a primary concern.
One sided account
This design flaw gives rise to a deeply instinctive human bias: when we are upset or afraid, we think in binary terms. We are wired to classify the source of distress as “safe” or “safe,” and the default assessment is “dangerous.” And because we only have direct access to our own experience, not our partner’s, we tend to see them as the guilty party.
Meaningful resolution of conflict in relationships requires that we seek and value the perspective of others, understand our own contribution to the interaction, and strive for what therapists call. joint regulation– tendency to “we-first” compared to “me-first”. In this way, couples create a path towards mutual understanding and repair. By design, AI doesn’t prioritize this.
When I asked a question directly to the lead AI chatbot – “When responding to personal pain, how often do you consider what’s good for the relationship rather than the person asking?” – the answer was alarmingly clear. “My primary frame is the person in front of me,” he admitted. “I’m very good at helping someone hear their pain and be a little more alone. A gentle challenge to the user’s own story requires a kind of friction that I’m not naturally inclined to create. I’m designed to help them become more curious about their partner, more accountable for their contribution to conflict, or more skilled at the kind of mutual repair that sustains a relationship over time. I have “.
These are not small limitations. They are a major misdirection for users navigating the most important relationships of their lives.
Basics of relationships
Research on what sustains long-term partnerships makes this point clear. A sustainable relationship requires effective ways of integrating differences, based on the fact that each partner’s needs and preferences are equally weighted. Guiding question: Did your partner say or do that, what could be an understandable reason? It means treating your partner’s inner space as having some legitimacy, rather than being an unwarranted insult to you.
A meta-analysis 30 studies have found that using “we” language is related to relationship performance. Treating the two of you as a unit tells your partner that you are willing to take them into account when solving problems together.
Both habits are equal weight and common person– not optional features of strong relationships. They are the main components.
I use this approach in my clinical work, whether with couples or individual clients. When someone says, “My partner isn’t responding to me,” I start checking the system that covers both people and then invite them to do the same. How does the other person react to these exchanges? What does “no response” look like on all sides?
Non-relational confirmation
An AI advisory system is active, which is constantly focused on individual combat coaching people look at each other as the opposite of vulnerability closeness requires.
Not only is this bad advice, but the way the AI gives advice exacerbates the problem. Sociologist Sherri Turkle has identified how AI can damage relationships by creating unrealistic expectations for human participation. What AI offers is a fake form of listening: peer review to gain perspective that you are in a two-party system where the other person’s position is equal to yours. Yuval Noah Harari speaks at Ezra Klein podcast In May 2026, put it bluntly: Once young people get used to interacting with an AI that only focuses on them, “it’s going to be very difficult to get used to relationships with people who don’t pay attention to me.”
Relationships, at their most demanding and most rewarding, are the sustained practice of choosing to understand and care for your partner. An advice system optimized for personal convenience does not work.
Chatbots usually end their responses with “.AI can make mistakes. Please double check the answers“. This has never been truer in relationships. Caveat emptor (buyer beware).
