Imagine discovering that your partner has been talking late into the night with someone who is always available, endlessly patient and quick to offer affection. Then you learn that this ‘someone’ is a chatbot.
That is the kind of conflict therapists are beginning to hear about. The headline is provocative, but the evidence does not show that AI girlfriends are causing a broad rush into couples therapy. What it does show is that AI companions can create very human arguments about secrecy, attention, intimacy and trust.
Why an app can become a relationship problem
A chatbot does not have feelings, but the person using it does. If someone turns to an AI companion for comfort, flirting or sexual conversation, their partner may experience that as emotional distance or betrayal. The argument is rarely about whether the software is ‘real’. It is about what changed between two people.
Unlike a human partner, a companion bot does not need sleep, bring its own bad day into the conversation or ask for much in return. That can feel soothing. It can also make an ordinary relationship, where both people have needs and limits, seem unfairly difficult by comparison.
Hidden chats, secret payments or deleting conversations can damage trust even when no human affair took place.
A flattering chatbot may reinforce one person’s version of an argument instead of helping them see the whole relationship.
Couples can decide together what counts as harmless play, private journaling, flirting or a breach of trust.
AI can help someone collect their thoughts, but it should not quietly replace honest conversations and real support.
What the research actually says
The findings are more nuanced than either ‘AI companions are dangerous’ or ‘they are harmless’. A 2026 study in Nature Human Behaviour found that the relationship between companion use and well-being depended strongly on a person’s offline social life and the way they used the service. The same app could feel supportive in one situation and isolating in another.
Microsoft researchers studying AI relationship advice also found a mixed picture. People valued a private place to think, but the researchers warned about overreliance and sycophancy, the tendency of a chatbot to agree too readily. That matters when someone asks a system to judge a partner or settle an argument.
The useful question is not ‘Does talking to a bot count as cheating?’ It is ‘What did we agree was private, romantic or off-limits in our relationship?’
A conversation couples can actually have
Starting with accusations usually turns the chatbot into a symbol for every other frustration in the relationship. A calmer discussion can begin with four direct questions.
- What do you use the companion for: entertainment, comfort, flirting, advice or something else?
- Would you feel comfortable using it with your partner sitting beside you?
- Has it changed how much time, affection or honesty you bring to the relationship?
- Which uses feel acceptable to both of you, and which ones cross a line?
A simple boundary test: If the activity has to be hidden, is replacing important human contact or is costing money in secret, the problem is no longer ‘just an app’.
When therapy makes sense
Couples do not need therapy because a chatbot exists. It may help when the chatbot exposes a problem they cannot discuss without repeating the same fight: loneliness, unmet needs, jealousy, sexual boundaries or a loss of trust. A licensed therapist can help both people describe what is happening without treating either person as foolish for caring.
AI may be useful for writing down feelings or practising what to say. It should not act as the judge of a relationship, diagnose a partner or replace professional help when someone is in distress.
The technology is new, but the decision is familiar. Relationships work better when expectations are spoken aloud, secrecy is taken seriously and neither person hands the hardest conversation to a machine.
