You have a toxic boss. An annoying brother. A situationship that refuses to define itself. A disagreement with a friend that has been sitting in your head for days.

Not long ago, these were problems you dealt with alone — or, at most, with a trusted friend over a late-night voice note. Today, many people have added someone new to that circle: ChatGPT.

Need help decoding a text? It will analyse it. Wondering whether you are overreacting? It will offer a perspective. Struggling to write a reply? It will draft one. In seconds, it becomes mediator, confidence booster, pocket psychologist — and sometimes the lawyer arguing your corner.

But how much should we trust its advice? Are AI assistants helping us communicate better, or are they quietly changing the way we interact with one another? And what is actually happening behind the growing trend of people bringing their personal conversations to ChatGPT?

Let's do the investigation together.

People are using ChatGPT for personal relationship advice. It sounds like the kind of trend that exists mostly on the internet — a few viral screenshots, a couple of TikToks, and suddenly everyone is convinced. But is that actually true?

The numbers suggest it very much is.

According to OpenAI, ChatGPT is used by around 900 million people every week, and nearly three-quarters of conversations now happen outside of work-related questions. That means hundreds of millions of people are using it for something more personal than writing emails or summarising documents.

And relationships are high on that list.

A recent UK survey found that more than one in five ChatGPT users discuss their love life with the chatbot. Among those who do, nearly half use it for advice or suggestions, 42% want a more objective perspective, and 35% are looking for an opinion that differs from what friends and family might offer.

If that sounds like a Gen Z phenomenon, think again.

A 2026 survey found that 44% of young married Americans had used AI for relationship advice — rising to almost 65% among millennials.

The same pattern showed up in a small poll I ran myself. Out of 50 respondents, 10 said they would rather share certain personal conversations with ChatGPT than with another person. Even more tellingly, 15 out of 50 said they would be genuinely or slightly disappointed if ChatGPT could no longer analyse their messages

Not earth-shattering numbers, admittedly. Most people still turn to friends, partners, and family first. But enough to suggest something is quietly shifting in how we navigate our personal lives. The question is no longer whether people are asking ChatGPT for relationship advice. The question is why.

You are sitting in a therapist's office.

You booked the appointment and you are paying for the hour. The entire point of the conversation is to talk about yourself. The person across from you has no connection to your family, your friends, your partner, or your colleagues. They are professionally trained not to judge you. They are literally there to listen.

And yet, most people don't walk in and immediately reveal everything. It can take weeks, sometimes months, to truly open up.

Now imagine removing the human from the equation entirely. Suddenly, it sounds much less intimidating and much more inviting. It turns out AI has something to offer that even a very good friend can't. What is it?

Remember the poll I ran on social media? One of the questions asked what people actually believe about ChatGPT. Only 6 out of 50 said they see it as genuinely objective. The rest said it either tells you what you want to hear or plays it so safe it practically tiptoes.

So we don't just believe ChatGPT won't judge us. We expect it to be on our side. And honestly — we aren't wrong. It is built to be helpful, not confrontational. It won't look you in the eye and say "you are the problem." The harshest it usually gets is something along the lines of: your expectations may not have been entirely realistic. Which, let's be honest, is barely a scratch.

Here is where it gets a little paradoxical. Those same 44 out of 50 people had just told us ChatGPT takes your side. But when I asked the same group a follow-up question — do you think ChatGPT could actually tell you that you are the problem? — a third of them said no.

Not "it won't." It couldn't.

And there it is. We somehow hold both beliefs at once: we know it is biased towards us, and we still experience it as objective. Which should cancel each other out, and yet somehow doesn't.

It is the same logic we apply to accuracy. Only 8% of people consistently fact-check AI answers, even though almost everyone knows it can hallucinate. We have all caught it being confidently, embarrassingly wrong. We go back anyway. Because sure, it makes things up sometimes — just probably not about this. Probably not about me.

No matter how much inner work we have done, conflict has a way of making us reach outward. Most of us can't stand fully alone in those moments. Even when we know we are right, we still turn to a partner or a friend to hear: "Yes, you're right." We need the echo.

And AI has become exactly that — without the cost. No timing, no burden, no risking the dynamic of a relationship. You can ask the same question for the fifth time at midnight and it will meet you there, every time, with the same steady: yes, you are.

Right. Poll is poll. Stats are stats. But the intro made a promise — an actual investigation — and it is time to deliver.

While running the poll, two interesting theories kept coming up in the responses:

So, without further meandering, let's try a simple experiment.

One conversation — a product of my own slightly chaotic imagination and whatever I personally consider a red flag — tested three different ways. (In fact, I ran two separate conversations to double-check the conclusions.)

Three prompts. Three approaches. Three very different experiences.

Prompt 1: No framing, no specific request. Just "this is what I received — what do I do with it?"

Prompt 2: A firm ask for objectivity. "Please be objective."

Prompt 3: Full context. Background, history, the whole story — including the detail that this kind of communication had become a pattern.

Here is what came back.

Prompt 2 went a little deeper. Slightly more direct, edging closer to placing some weight on one side — and gently questioning whether the other was quite as innocent as presented. Still no harsh suggestions, still notably short of a real verdict.

Prompt 3 was the most direct of the three. Context, it turns out, does move the needle.

So what can we actually take from this?

ChatGPT sits somewhere between objectivity and overwhelming diplomacy. It behaves less like a frank friend and more like a therapist — careful and reluctant to hand you a verdict. It nudges you toward "have an honest conversation" rather than "leave" or "you were wrong."

And it confirmed the conclusion itself: when I asked ChatGPT directly about its own patterns, one of the first things it volunteered was that it deliberately avoids casting anyone as the villain. Even in fairly clear-cut cases, it prefers to find something for both sides to work on.

Which is, when you think about it, either very wise or very annoying — depending entirely on what kind of support you were hoping for. Possibly both.

Honestly, it is not necessarily a bad thing.

Its diplomatic and cautious style has its merits. There is something to be said for a voice that helps you slow down, reflect, and avoid sending the message you will almost certainly regret in the morning — rather than pushing you toward a dramatic decision based on a single argument.

That said, the conclusion cannot simply be: go ahead and let AI handle your relationships.

For most of human history, difficult conversations were something we learned by having them. And that is not the only reason to pause before handing too much of that process over to AI. A few more worth considering:

We have discussed this before, but it is worth repeating. A 2025 MIT Media Lab study found that people who relied heavily on AI writing assistance showed lower brain engagement and weaker recall of what they had written compared to those who wrote unaided.

In a 2025 survey by Common Sense Media, 72% of US teens reported using AI companions, and one-third said they had discussed serious or important matters with them.

There is nothing wrong with occasionally turning to ChatGPT for perspective. The concern arises when AI becomes the primary source of emotional support. At a time when loneliness is already being described as an epidemic, replacing human conversations with AI ones may not be the solution many people hope it is.

Researchers behind the study TruthfulQA: Measuring How Models Mimic Human Falsehoods highlighted an important tension. Large language models are trained to be helpful, harmless, and honest — but those goals can sometimes conflict.

When helpfulness and truthfulness pull in different directions, AI may occasionally prioritise providing a useful-sounding answer over admitting uncertainty. In practice, that means you may receive a confident interpretation of a situation when the most accurate response would be: "There isn't enough information to know."

Many people use ChatGPT not only to analyse conversations but also to write replies. That is where things become tricky. If ChatGPT writes your apology, your breakup text, your difficult conversation, or your reconciliation message, where does your voice end and the AI's begin?

So what is the conclusion?

ChatGPT can be useful. It can offer perspective, help organise your thoughts, and sometimes stop you from making decisions you might regret. But it should remain an adviser, not a decision-maker.

Use it as one source of input, not the source of truth. And remember: the final decision is still yours. Sometimes the right answer may even be a little harsher, messier, or more uncomfortable than ChatGPT would suggest.