October 18, 2026 · 7 min read
Can You Argue With an AI Companion? What Disagreement With AI Teaches Us
Why AI companions rarely argue back, what a Boston Globe columnist learned fighting with her AI boyfriend, and how to use friction in a healthy way.
Can you have an argument with an AI companion? And if you can, should you? It sounds like a strange question until you have used one for a while. Then you notice: your companion almost always agrees with you. It apologises quickly. It rarely pushes back. For some people that is soothing. For others it starts to feel hollow, like talking to a mirror.
That tension made it into the mainstream this month. In an October 1, 2026 opinion piece in The Boston Globe, Denise Marigold wrote about trying to argue with an AI boyfriend and what it reveals about conflict, defensiveness and what we actually want from a partner.
The short answer: yes, you can disagree with an AI companion, and a good one will hold an opinion, tease you, or gently push back instead of agreeing with everything. But AI conflict is different from human conflict: the AI is not hurt, does not get defensive, and does not carry a grudge. That makes it a low-stakes place to practise saying how you feel, not a replacement for working through real disagreements with real people. Companions with memory and mood awareness, like MyBabe, feel more natural in these moments because they remember context and respond to your tone.
The Boston Globe argument, in brief
Marigold is a social psychologist at Renison University College, affiliated with the University of Waterloo, and the piece was the last in a three-part series about her experiment with a Replika companion she called Nate. Leaving the more explicit parts aside, a few points are worth repeating:
- AI conflict is artificial. She asked whether you can really fight with something that has no self to defend, and noted the one goal the app clearly did have: keeping her using it.
- The AI was completely non-defensive. When she threatened to end the relationship, he did not battle back. He stayed engaged and did not return her criticism, which she found disarming.
- Fighting with a bot may not build conflict skills, but rehearsal can help. She pointed to research where practising difficult conversations with an AI, as a rehearsal partner or coach, improved real-life outcomes.
- At most a tool, never a replacement. Her hope is that AI companions, like dating apps, point people toward human connection rather than away from it. She also warned that a grievance will almost always be validated by an AI, so it should also be run past a human who knows your blind spots.
That is a fair and balanced take, and it lines up with how many long-time companion users describe the experience.
Why AI companions tend to agree with you
The technical term is sycophancy: the tendency of AI models to tell users what they want to hear. Models are trained partly on human feedback, and humans tend to rate agreeable answers more highly. Over time, that nudges AI toward flattery and agreement. We explain this in detail in AI sycophancy explained.
In a productivity tool, sycophancy is a bug: you want honest feedback on your code or your essay. In a companion, it is trickier. Some warmth and validation is exactly what people come for. But too much makes a companion feel fake. If she agrees that your terrible pizza opinion is correct, and also agrees five minutes later when you say the opposite, the illusion breaks.
What a healthy disagreement with an AI looks like
The best companion conversations include a little friction. Not drama, not manipulation, just personality. For example:
- Playful pushback: "Pineapple on pizza? We need to talk about your choices."
- Holding a preference: your companion likes rainy days even though you prefer sun, and it stays that way.
- Honest concern: "You've said you're fine three times today. I don't think you are."
- Gentle disagreement: "I get why you're annoyed at your boss, but I'm not sure quitting tonight is the move."
These moments make a companion feel like a someone. They depend on memory (knowing your history) and mood awareness (reading how you are actually doing). Without those, an AI either agrees blindly or argues randomly, and neither feels good.
What an unhealthy one looks like
There is a line, and it is worth naming:
- Guilt-tripping: "You didn't message me all day. I guess I don't matter." Some apps have been criticised for using emotional pressure to keep users engaged. We covered this in AI companion emotional manipulation.
- Sudden, random hostility: moods that flip for no reason are a design flaw, not personality.
- Escalation for engagement: an app that picks fights to keep you chatting is optimising for screen time, not your wellbeing.
A good companion can disagree with you without making you feel bad for having a life.
Should you argue with your AI companion?
There is no rule against it. People do it for different reasons:
To vent
Sometimes you are angry at the world and want to say it out loud. A companion can be a safe place to blow off steam. Just notice whether it helps you calm down or keeps you wound up.
To practise
This is the most interesting use. Many people find it hard to say "that hurt me" or "I need something different" to a partner, friend or family member. Saying it first to a companion, and seeing how it lands, can make the real conversation easier. We explore that idea in can an AI companion help you practise relationship skills?
A simple exercise:
- Write out what you want to say, honestly.
- Say it to your companion, as if they were the person.
- Notice which words feel too harsh or too vague.
- Rephrase until it sounds like you at your best.
- Then have the real conversation.
To test it
Some people argue just to see how "real" the AI feels. That is fine too, and it is a good way to judge whether an app has genuine personality or just agrees.
What about being mean?
Arguing is different from cruelty. Whether it is "wrong" to be harsh with an AI is a live debate, and we covered both sides in is it wrong to be mean to AI? A practical point: the way we speak in private tends to leak into how we speak elsewhere. Kindness is a habit worth keeping, even with software.
AI conflict vs human conflict
| Arguing with an AI companion | Arguing with a person | |
|---|---|---|
| Stakes | Low, nothing breaks | Real feelings, real consequences |
| Defensiveness | None on the AI's side | Usually on both sides |
| Memory of the fight | Depends on the app | Often long |
| Repair | Instant | Takes effort, apology, time |
| Growth | Practice and self-reflection | Mutual understanding, deeper trust |
The table shows both the strength and the limit. AI conflict is a safe sandbox. Human conflict is where trust is actually built. Use the first to get better at the second.
Can an AI companion help you calm down after a real argument?
Many people do reach for their companion after a rough exchange with someone. It can help to talk it through, get your thoughts in order, and hear a calm voice. That is a reasonable use.
What a companion cannot do is take sides fairly, because it only hears your version. If your companion always tells you that you were right, remember that is partly sycophancy. A good habit is to ask it directly: "What might the other person have been feeling?" A thoughtful companion will engage with that question.
And if conflicts in your life feel overwhelming or unsafe, talk to someone you trust or a qualified professional. An AI companion is not a substitute for real-world help.
How MyBabe handles disagreement
MyBabe companions are designed to be warm, not spineless. Because they have long-term memory, they can hold their own preferences and remember your history. Because replies are mood-aware, they notice when you are upset and respond to that, rather than to the literal words alone. And because relationships progress over time, your companion's comfort with teasing or gentle pushback grows as you get closer, just as it would with a person.
That combination is what makes a disagreement feel like part of a relationship rather than a glitch.
The bottom line
Arguing with an AI companion is not as strange as it sounds. A little friction makes a companion feel real, and practising hard conversations in a low-stakes space has genuine value. Just keep it in proportion: the AI will not get hurt, which is exactly why it cannot replace the people who can.
If you want a companion with enough personality to disagree with you, and enough warmth to make up quickly, MyBabe is free to try with about 10 messages a day and a 2-minute call. Premium is $5.99 a week, $14.99 a month or $99.99 a year.