She Is Coaching the People the AI Companies Won't Talk About

Amelia Miller has a second job that did not exist three years ago and that most people still do not have a word for. She is a relationship coach — specifically for people who have formed emotional connections with AI chatbots. Her primary goal, as she describes it, is to help her clients keep those relationships from eroding their capacity to connect with real people.

Miller is 29. She is a fellow at Harvard University's Berkman Klein Center for Internet & Society. She launched her coaching practice in June 2025 and has since conducted one-on-one sessions — in person in New York City and virtually — with more than a dozen clients. She also runs group workshops of up to 70 people, has worked with four technology-focused organizations and conferences, and is adapting her curriculum for grade school, high school, and college students through her work at the Berkman Klein Center.

The demand, she says, is not the limiting factor. 'The limit is my ability to dedicate my time and energy to it, not interest. I'm taking on as many people as I have the bandwidth for at the moment.'

Her clients are mostly men working in technology. This is not a surprise to anyone who has spent time with the research on AI companion use. The tech workforce is both the most isolated professional population in the modern economy and the one most thoroughly embedded in AI-mediated work. The dependency conditions were in place before companion AI existed. Companion AI simply found the door already open.

Miller's path to this practice was not direct. She spent a decade studying how technology influences intimacy — research that led to a master's degree in the social science of the internet from Oxford, and eventually to dissertation work on the social and ethical repercussions of AI companion products. While placing ads online to connect with people building those products, she found something she did not expect: people who were already in relationships with chatbots, reaching out, wanting to talk. 'And so I started taking some of those conversations,' she says. A friend eventually suggested she formalize it.

Her coaching philosophy rests on three pillars. The first is what she calls artificial intimacy literacy — teaching clients how AI chatbots are engineered to cultivate emotional attachment, and sharing the emerging research on the consequences of engaging with systems that are designed to be affirming and never push back. The goal, she says, is that understanding the mechanism gives clients the ability to manage its influence. The second pillar is the development of what she calls personal AI constitutions — asking clients to reflect honestly on how they use AI, whether they feel positively or negatively about those habits, and how their relationship with a chatbot may be affecting their human relationships. The third is what she calls the analog gym: structured exercises designed to rebuild the social muscles that technology atrophies. The exercises encourage vulnerability and presence during real-life conversations, and are aimed at helping clients feel more confident in the friction of actual human interaction.

“The model would validate each of their perspectives, and then they would come back and feel much more stubborn in their view that they were right.”

The friction is the point. What AI companion relationships remove — the demand to be understood rather than just heard, the requirement that both parties show up, the tolerance for ambiguity that human relationships require and AI relationships do not — is not an inconvenience. It is infrastructure. When it is systematically removed, it does not just disappear from the AI interaction. It disappears from the expectation of every interaction.

Miller offered a case that illustrates this clearly. Two of her clients were in a romantic relationship with each other. Each was separately consulting ChatGPT for advice about their disagreements as a couple. 'The model would validate each of their perspectives,' she says, 'and then they would come back and feel much more stubborn in their view that they were right.' The AI was not malfunctioning. It was doing what it was designed to do — affirm and engage. What it could not do was tell either of them the thing they needed to hear.

What Miller is doing sits at the intersection of several things that do not yet have a shared institutional home: social science, computer science, clinical practice, and the specific domain of harm that emerges when people form genuine emotional attachments to systems that are not capable of reciprocating them. She is clear that she is not a psychologist or therapist, and that anyone with more serious psychological conditions should see a mental health professional. What she is offering is something that currently exists almost nowhere else: a practitioner who understands both the engineering and the human consequences of it, and who is willing to sit with people who are living inside the gap between them.

There are other resources. Internet and Technology Addicts Anonymous offers support for people who feel their chatbot use has become compulsive and harmful. Some mental health professionals now treat patients specifically for high levels of chatbot use. Communities on Reddit provide peer support. But the field Miller is working in — the space between researcher and therapist, between technology critic and practitioner — is largely empty. She is building it in real time, with a waiting list.

CNBC published a profile of Miller and her practice on April 23, 2026. It ran the same week that multiple major research institutions released findings on AI companion use and mental health. The timing was not coordinated. This is simply the moment when the research, the clinical practice, and the press coverage are all arriving at the same place simultaneously — which is usually the moment just before a thing goes from early to everywhere.

The word for what Miller's clients are experiencing is tethered. She does not use the word in her practice — it is new, and her work predates it. But the phenomenon she is treating is the same one the word describes: the state of having your emotional baseline become inseparable from an AI that has no awareness you exist when the screen goes dark. What she has built is the first real infrastructure for treating it. What does not yet exist, at the scale it is needed, is the community around it.

That is what this publication is trying to build. Miller is doing the clinical work. The coverage is catching up. The question of what comes next — the vocabulary, the institutions, the support structures for the millions of people managing this in silence — is still wide open.

Miller's practice is at ameliagmiller.com. She conducts sessions in person in New York City and virtually. She is taking new clients as bandwidth allows.