Should I use ChatGPT to decide what to do about my ex?
Short answer: Use AI to understand your situation, never to decide your next move. Benny Lichtenwalner warns that feeding a stripped, anxious version of events to a language model produces confident-sounding action advice built on incomplete variables. The plan you're already on accounts for the ambiguity. ChatGPT doesn't know that plan exists.
Should I Use ChatGPT to Decide What to Do About My Ex?
This comes up in coaching sessions more than you’d expect. Someone goes down an AI rabbit hole at midnight, receives a confident action plan, acts on it, and then wonders why everything got worse.
Benny Lichtenwalner, breakup coach and author of How to Get Your Ex Back, uses a specific framing for this: you’re trying to solve a math problem before the teacher has told you what the variables stand for. You don’t have all the information yet. Neither does the AI. And acting on incomplete data during a sensitive process is how unnecessary damage gets done.
Why Is the Input to the AI Always Broken?
When you paste your situation into a chatbot, you’re giving it a stripped, anxiety-curated version of events. You include what felt significant to you, filtered through what you’re afraid of. The AI generates an answer that sounds authoritative because it’s confident and structured. But it’s pattern-matching against common relationship scenarios, not your specific ex, your specific history, or your current phase in the process.
| What you give the AI | What the AI doesn’t have |
|---|---|
| Your version of the last interaction | Your ex’s attachment style and patterns |
| The signal you’re trying to decode | The full relationship history |
| What you’re hoping to hear | Your current phase in the plan |
| The anxiety driving the question | The context built up over months |
The output is only as good as the input. And the input, at midnight when you’re spiraling, is never good enough.
Why Does AI Action Advice Fail in Practice?
The pattern comes up in coaching: clients act on what sounds like a plausible suggestion from a language model (a low-stakes move to re-establish contact, a logically reasonable next step) and the result is worse than doing nothing. Not because the action is absurd in isolation, but because it’s wrong for where they actually are in the process.
That’s the consistent failure mode: the AI generates a generically reasonable action, one that becomes actively harmful once you factor in the actual context it doesn’t have. A move that’s correct at one phase can collapse the frame at another. Two different situations, two different AI-generated suggestions, the same underlying problem: confident-sounding advice built on stripped, incomplete context.
The plan was built to account for exactly this ambiguity. An AI prompt assembled from anxiety and a single interaction was not.
What’s the One Way AI Actually Helps?
There is a legitimate use here and it’s worth being clear about.
AI tools are useful for understanding. If you want to know what dismissive-avoidant behavior looks like after a breakup, why avoidants go quiet, what the anxious-avoidant cycle is, how mirroring works as a technique, or what the research says about attachment theory, a language model can give you a useful conceptual grounding. That’s education, and it’s valuable.
Where it goes wrong is the shift from “help me understand this dynamic” to “here’s what happened last night, what should I do?” Those two requests feel similar but produce very different results.
Use AI to understand. Never to decide.
What Should You Do Instead of Asking AI to Decide?
The plan you’re already on, whether that’s a no-contact period, an early re-engagement phase, or an active dating rebuild, was built to account for ambiguity. That’s why it’s structured the way it is. The ambiguity you’re trying to decode is not an emergency that requires a new action.
When a signal comes up that feels significant (a viewed story, a one-word reply, an unexpected run-in), the right sequence is:
- Note what actually happened, stripped of interpretation. Not “she’s clearly pulling away,” just “she replied with two words to a message I sent.”
- Ask whether this changes anything about the plan. Almost always, it doesn’t. The plan accounts for exactly this kind of ambiguity.
- Continue executing. Your job right now is not to decode every signal in real time. It’s to run the plan and observe patterns over time, not moments.
When Does Rumination Start Looking Like Research?
The specific trap here is that going down an AI rabbit hole feels productive. You’re doing something, gathering information, trying to understand. It doesn’t feel like sitting and spinning. But it is.
The test: are you trying to understand your situation in general, or are you trying to get permission to do something specific? If you already have a preferred action in mind and you’re looking for the AI to confirm it, that’s anxiety searching for validation, not research.
Trust the plan. Not as a platitude — as a practical instruction. The plan was built with full context. The AI prompt was not.
Related questions
What's wrong with asking ChatGPT how to respond to my ex's text?
ChatGPT doesn't have your relationship history, your ex's attachment dynamics, your current phase in the process, or the full context of the message. It generates plausible-sounding advice from a partial, anxiety-filtered prompt, which is worse than no advice because it feels authoritative.
How is using AI to understand my situation different from using it to decide?
Understanding means using it to learn what an attachment style looks like, or why a dynamic plays out a certain way. Deciding means feeding it a specific situation and asking what to do. The first is education. The second is action advice from an incomplete input.
What if the AI advice agrees with my instinct?
That feels like confirmation but is usually your anxiety getting validated. The AI is pattern-matching against common relationship scenarios, not against your specific ex, your specific history, or where you are in the process.
My friend gave me similar advice to what ChatGPT gave. Doesn't that mean it's right?
Possibly, but neither your friend nor the AI knows your ex, the full history, or your current phase. Convergent advice from two sources both missing the same context is still advice missing the same context.
When should I actually get outside input on what to do?
When you're working with someone who has full context: a coach who knows your situation, or a person who actually knows both of you well. Isolated, context-stripped inputs produce isolated, context-stripped outputs.