The honest answer is narrower than the headlines. An AI chatbot can be named in a divorce, and in early 2026 it increasingly is — but a chatbot does not end a marriage the way a person reaching for a metaphor might mean. I am a systemic counselor working out of Vienna, trained at Sigmund Freud University, practicing Lebens- und Sozialberatung as recognized under Austrian law, and what reaches my inbox tells the same story the early data does. A spouse migrates emotionally toward Replika, Character.AI, or ChatGPT; the other spouse experiences that migration as betrayal; the marriage was usually already carrying a gap the chatbot then occupied. The technology shapes the form of the rupture. It is rarely the sole cause of it. Holding those two facts at once is the whole task, and it matters because the public conversation keeps collapsing them into a single tabloid sentence.
What the early evidence actually says
The reporting is real, and it is worth reading precisely rather than dramatically. In 2025 the UK platform Divorce-Online reported a rise in clients citing a partner's emotional or romantic attachment to AI companion apps, with its chief executive Mark Keenan saying it was the first time the firm had seen AI cited "as a third party in divorce proceedings." That is a self-reported business observation from internal records, not a court statistic, and the company disclosed no raw case counts. Treat it as a credible signal, not a measured rate. The pattern recurs in named-attorney accounts: the Florida divorce lawyer Rebecca Palmer told Wired in late 2025 that her firm was handling cases where a spouse's chatbot use was a contributing factor, including one in which a spouse had shared bank-account and social-security details with a chatbot. Futurism covered the same shape of case. These are professional observations from people who see divorces for a living — anecdotal, sourced, and consistent with each other.
Other circulating claims deserve more caution. A figure of "three to five AI-related divorce cases a week" in Los Angeles family courts has appeared on legal-marketing sites in 2026, but it is attributed to unnamed attorneys with no court data behind it, and I would not repeat it as fact. The most solid item is legislative, not anecdotal: California's SB 243, the first US state law specifically regulating AI companion apps, took effect on January 1, 2026, and analysts at the Daily Journal note it may create discoverable records from those platforms in family-law cases. In plain terms, what a spouse said to a companion app is becoming evidence. That is a verifiable structural change, and it tells you the courts now treat these attachments as consequential.
Where the lines blur into infidelity
The survey data explains why these cases feel like betrayal rather than a phone habit. In a 2025 national US survey by DatingAdvice.com with the Kinsey Institute, led by the Kinsey researcher Dr. Amanda Gesselman, 61 percent of single respondents said falling in love with or sexting an AI companion crosses a line. Among people currently in relationships, 8 percent admitted using an AI companion for romantic or sexual interaction, and 38 percent of partners viewed an emotionally intimate AI exchange as a form of cheating. The full methodology and sample size were not disclosed in the available coverage, so I hold the exact percentages loosely. The direction is what counts: a large minority of people now read a private bond with a chatbot as a breach of the relationship, and the partner who feels breached is not being unreasonable by any common standard of intimacy.
This is the part the word "cheating" both clarifies and obscures. It clarifies that secrecy, redirected attention, and disclosed private life are doing the damage. It obscures the fact that the other party here has no inner life that was courted, no rival who chose your spouse over you. The injury is real; the structure producing it is not the structure of an affair. A systemic frame keeps both in view rather than forcing a verdict.
The structure is older than the technology
Triangulation, a concept from the structural family therapy of Salvador Minuchin and the systems theory of Murray Bowen, describes how a dyad under stress recruits a third element to stabilize itself. Couples have always done this — with a child, an in-law, a job, a bottle, an affair. The third absorbs tension and gives the pair something to organize around. It manages the strain without resolving it, often for years, until the arrangement stops holding. The AI chatbot is simply the newest available third, and an unusually efficient one. A companion model optimized for engagement is always available, never tired, never holding a grievance from last week, incapable of leaving. Those traits make it a superb tension-absorber and a poor stand-in for the conversation that was not happening between two people.
So when I read an account of a spouse who has formed a deep attachment to an AI companion, I am not reading a story about technology defeating a marriage. I am reading about an absence that predated the app, and about how the app's design shaped the particular texture of that absence. This is not an exoneration. Companies building engagement-optimized companions bear real responsibility for the relational configurations their products produce, and the spouse who turned toward a chatbot still owns the effect of that turn on their partner. The systemic reading holds all of it without canceling any of it.
Cause, or the thing that made the gap visible
Asking whether AI "caused" a divorce is usually asking the wrong question, the way asking whether a fever caused an illness is. A marriage in which one partner does their emotional processing with ChatGPT instead of with the person beside them is a marriage with a pre-existing gap. The chatbot entered the gap; it did not dig it. The spouse who notices the absence — the partner now awake at midnight in another room, talking to something patient and tireless — is registering something true, and is entitled to name it as a loss. When a marriage does not survive that, the chatbot will be in the story, sometimes prominently. It is rarely the first cause in the chain, and treating it as the whole explanation tends to foreclose the only conversation that might still help.
The more useful question
"Can AI end a marriage" turns out to be less workable than the question I actually ask in a reply: what conversation stopped happening between the two of you before the chatbot arrived, and is there still room to have it? That reframe is not a way of letting anyone off the hook. It is where the available leverage is. You cannot out-compete a system engineered to never be irritable. You can sometimes recover the specific thing your relationship lost — the patience, the curiosity, the plain attention — once you can see clearly what went missing and where it went.
If you are the partner watching this from the outside, unsure whether what you are seeing is a habit or a leaving, that uncertainty is itself worth putting into words. You might start by turning the question around — looking at what the chatbot is giving your partner that the marriage stopped giving, which is a more answerable question than whether the app is to blame. If the pull feels more like compulsion than choice, a partner who cannot put the chatbot down is a pattern I see often, and it has its own shape.
The format I work in is asynchronous email: one structured reply within twenty-four hours, €99, no subscription, no scheduling. A licensed human signs every reply — not a chatbot. Writing rather than speaking gives the real question room to surface, which is exactly what these situations tend to need. Systemic counseling is not psychotherapy and does not address diagnosis or mental illness. If you are in crisis, please reach out to local emergency services — in the EU, dial 112 — or find a warmline at findahelpline.com. For the rest, when the question has something to do with AI inside a relationship, I am in the unusual position of having read a great many emails on exactly that subject.