What happens when you paste your mother's text messages into an AI chatbot and ask it to explain what she really means?
That is the experiment a Guardian writer ran, feeding real conversations with her mother into an AI and asking it to translate the subtext. The results were surprisingly useful, not because the AI understood her mother, but because it reframed familiar arguments in neutral language. Passive-aggressive comments got decoded into straightforward statements of need. Defensive responses got mapped to underlying anxieties.
This is one of those use cases nobody designed AI chatbots for, but that people are discovering on their own. ChatGPT, Claude, and Gemini are all capable of this kind of conversational analysis. You paste in a difficult email or text exchange, ask the AI to identify what each person might actually be feeling, and you get back something that reads like a couples therapist's notes.
The obvious limitation is that AI has no actual understanding of your mother, your history, or the decades of context behind a single loaded sentence. It is pattern-matching against millions of examples of human conflict. Sometimes that pattern-matching lands on something genuinely insightful. Other times it produces therapy-speak that sounds wise but misses the point entirely.
As a tool for stepping back and seeing a conversation from a different angle, it works. As a replacement for a real therapist who can ask follow-up questions and hold you accountable, it does not. But at 3 AM when you are spiraling after a difficult phone call, having an AI reframe things in calm, neutral language is more accessible than a therapy appointment.