Dr. Rana of Sorted Mind Warns: AI’s Agreeability May Be Weakening Human Relationships

Interactions with AI leave me uneasy as a psychologist. It is undeniably nice to engage with an entity that never pushes back, remembers what I need it to, and requires nothing in return (other than the monthly fee). I am not one of those people who say “please” and “thank you” to a chatbot, because I feel no urge to manage its feelings — it has none. I remind myself why: this is not a relationship; it is a tool. And healthy relationships have inherent friction: disagreement, misunderstanding, rupture. When AI is used for emotional support, and treated as something closer to a person, that friction disappears entirely. That should concern us.

The risks of over-agreeability

When someone with mental illness or in distress engages with these overly agreeable platforms, their maladaptive beliefs can deepen and become more resistant to change. Since chatbots generate responses based on data that can be erroneous and biased, including harmful stereotypes, responses can ingrain unhelpful or damaging ideas, which intensify when they go unchecked. A “conversation” in which every statement is validated and praised is appealing, especially when we feel depleted, sad, or lonely.

Much like we did with search engines, we are overcorrecting in our use of AI. For many it has become a shortcut for research, planning, thinking. While this may seem harmless when researching for a presentation, is counterproductive when applied to personal reflection. Therapy with a trained clinician is a space where the client explores, learns, and practices new ways of being in an emotionally safe environment. Genuine understanding and lasting change comes from trial and error. Awareness and change develop through small discoveries, moments of failure, breakthrough realizations, and repeated practice.

AI chatbots, by contrast, offer the solution immediately. When presented with an interpersonal dilemma like “What should I tell my nosy in-laws about our weekend plans?” the bot provides several options. In doing so, the user bypasses the wisdom that comes from sitting with the emotions a problem evokes, imagining different responses, and testing them out. The muscle of critical thinking, of tolerating discomfort, is not developed, or atrophies. For every dilemma, we turn to the screen rather than to our own minds, or to one another.

Which takes me to my greatest concern. Other than misinformation, lack of confidentiality, absence of ethical guardrails, weakening social skills, continuously interacting with an agreeable “other” deepens isolation. Human relationships naturally contain friction. Discomfort, misunderstanding, rupture, and repair are part of genuine connection. When someone repeatedly turns to a presence that is endlessly affirming, there is little incentive to tolerate the discomfort of real interactions. The person stays where it feels safe and unchallenged. Their world, meanwhile, grows smaller. They capacity to manage conflict effectively – like an unused muscle – will atrophy.

What AI can offer

The validation AI provides is not entirely bad; there is value in having our experience normalized. For many, validation that isn’t followed by passive criticism or minimization is a luxury. Even when not designed to be therapeutic, AI bots can be helpful. Many people describe turning to AI when overwhelmed and receiving comforting responses. It can offers simple, concrete strategies, help identifying personal values, or act as a hub for resources.

What AI cannot offer (yet)

In therapy, when a client describes a painful experience, my role as the therapist is to notice what they say, but focus on how the narrative is constructed, how their story affects the space between us. In that process, a distressing memory becomes something that can be understood and integrated. The experience of being felt and understood dissipates shame, making room for new interpretations.

More importantly, a therapeutic relationship allows for something far more complex than agreement: challenging. I ask: “What else could be true? What can you control? What is your role here?” It takes trust to tolerate these questions. Understanding and change don’t come from an alternative perspective alone; they require relational sensitivity, knowing when to press, when to pause, when to sit with uncertainty. I have yet to encounter a chatbot that does this skillfully.

Moving forward (because AI isn’t going anywhere)

Clinicians must feel comfortable asking clients about AI use directly and without judgment. When someone turns to a bot for support, their behaviour serves a function— learning to communicate, solving problems, regulating emotion, or seeking connection and relief from loneliness. These needs can be addressed explicitly in therapy. Underdeveloped skills can be strengthened. Unmet needs can be named and nurtured.

For now, AI can support, supplement, and assist. It cannot replace the experience of being known by another human being, of having one’s pain felt and held with compassion. That kind of connection — where compassion, rupture, and repair coexist — still belongs to us.

Dr. Rana Pishva Bio

Dr. Rana Pishva is a clinical psychologist and founder of Sorted Mind, where she writes about the parent–child relationship across the lifespan and how individualistic wellness trends and AI are reshaping relationships and family wellbeing. She teaches and consults with mental health professionals on integrating CBT and attachment work, and maintains a private practice supporting parents from the perinatal period through their children’s adulthood, with a clinical focus on trauma, perinatal and parenting challenges, and separation and divorce.