When someone writes to me from Vienna, or Zurich, or Melbourne, about a partner's AI use, one thing recurs before the facts do: a belief stated as if it were already settled. He prefers it to me. She trusts a chatbot more than she trusts me. This means our relationship is over. The sentence arrives with the texture of certainty — and the certainty is doing most of the damage. I am Stefan Kohlweg, a licensed systemic counselor working through Lebens- und Sozialberatung in Vienna, trained at Sigmund Freud University, and what I notice across these cases is that the belief itself, not the AI's actual role, is often what tips a difficult situation into a crisis. Byron Katie's method, which she calls The Work, is the most direct tool I know for loosening that grip. It does not require a therapist. It requires a pencil and forty minutes of quiet honesty.
What The Work is — and what it is not
Byron Katie developed The Work in the 1980s in Barstow, California, following a period of severe depression and a sudden shift in her experience of suffering. She described the shift as waking up to the realization that her thoughts about the world were causing her pain, not the world itself. The method she built from that realization is disarmingly simple: write down a stressful thought, then interrogate it with four questions, then turn the thought around. The Work is not positive thinking. It does not ask you to replace a negative belief with a positive one. It asks you to question whether the belief is actually true.
In the context of AI relationships — situations where one partner has formed an intense connection to ChatGPT, Claude, Replika, or a similar system — the charged beliefs tend to cluster around replacement, preference, and betrayal. These are real emotions pointing at real dynamics. The Work does not dismiss them. It asks whether the story you are telling about those dynamics is the most accurate one available, or whether it is one among several possible readings.
The four questions
Take a specific, concrete belief you are holding right now. Not a general complaint — a particular sentence with a subject and a predicate. Something like: My partner uses AI because talking to me is not enough. Or: She tells the chatbot things she will never tell me. Hold that sentence in mind, and then move through these four questions slowly, one at a time.
Byron Katie's four questions
- Is it true?
- Can you absolutely know that it's true?
- How do you react — what happens — when you believe that thought?
- Who would you be without the thought?
The first question is the easiest to rush past. Most people answer yes and move on, which defeats the process. The second question slows that down: can you absolutely know it is true? This is not asking whether you have reason to suspect it, or whether it feels probable. It is asking whether you have certainty — the kind that would hold up if your partner were in the room giving their account. Almost never. The gap between "I believe this" and "I can absolutely know this" is the gap The Work is designed to examine.
The third question is where most of the useful material surfaces. When you believe my partner uses AI because talking to me is not enough, what happens in your body? What do you say? How do you treat your partner? How do you treat yourself? Most people find, when they sit with this, that the belief produces the very behaviors that make connection less available — withdrawal, accusations, a guardedness that confirms the story. The belief is not neutral. It has consequences.
The fourth question asks you to imagine — not as a permanent state, but as a brief experiment — who you would be in that same situation, with that same partner, without the thought. What would you notice that you currently cannot? What would you be able to say? This question is sometimes experienced as threatening, because the belief has organized the situation and removing it feels like losing ground. That feeling is worth noticing.
The turnaround
After the four questions comes the turnaround: taking the original belief and reversing it in as many directions as apply. My partner uses AI because talking to me is not enough can turn around to I use AI because talking to myself is not enough — meaning, my own internal narrative is insufficient, which is what brought me to this conversation in the first place. It can also turn to My partner uses AI because talking to them is not enough for me — which opens a different question about what I am not saying in our actual conversations. Or: My partner does not use AI because talking to me is not enough — which asks whether there are other plausible explanations I have not considered.
The turnaround is not meant to be believed wholesale. It is meant to be tried on, examined for whether it contains even a grain of truth. If the turnaround I use AI because talking to myself is not enough produces a flicker of recognition — if you have found yourself scrolling through a chatbot conversation to avoid sitting with your own thoughts — that is information. It does not exonerate your partner. It opens the field.
What this is for
The Work is a self-inquiry practice. It was never designed for couples work, and doing it on your partner's beliefs, rather than your own, is a well-documented misuse that tends to produce resentment. What it is designed for is examining the thoughts that have you — the ones that feel like facts rather than interpretations. In AI relationship distress, those thoughts often arrive pre-hardened: the situation has been mulled over so many times that the belief has acquired the texture of observation. The Work interrupts that.
It also has a structural advantage in this particular terrain. Katie's method assumes that suffering is generated by unquestioned thinking, not by circumstances. Whether or not you accept that framing fully, the operational point holds: the moment before escalation — before you send the confrontational message, before the argument that ends the evening — is almost always a moment in which a belief has been taken as given. Questioning it at that moment does not require any participation from your partner. It requires only that you be willing to not yet know.
A free first step you can take today
If you have been circling a belief about your partner's AI use — about what it means, about where it is leading — the Inversions tool on this site works in a similar register: it takes the situation as you describe it and offers the turnarounds, the alternative readings, the questions that tend not to surface when you are inside the feeling. It is free, takes about fifteen minutes, and a licensed human signs every reply — not a chatbot. You can also go directly to the situation page for a partner's heavy AI use if you want to read how others have framed the same ground.
If the situation has moved past a belief you can question alone — if the conversations are already happening, the damage is already visible, and what you need is a structured reply from someone who has read a great many emails on exactly this subject — that is what the counseling session is for.
If you are in acute distress or danger right now, please contact emergency services (112 in the EU) or find a helpline at findahelpline.com.