I Talked to ChatGPT About My Feelings Every Day, Until.

By Amaya Perera

There was a time when talking to a machine about your feelings would have sounded faintly absurd. You might have imagined someone sitting alone at a computer, typing their problems into a blank screen and waiting for a robot to tell them everything was going to be all right. Today, the image is considerably less strange. Millions of people have discovered that there is something remarkably comforting about an intelligence that is always available, never visibly impatient and capable of turning the tangled interior of a human mind into articulate, reassuring sentences. You can tell it what happened. You can tell it what you think happened. You can tell it what you wish had happened. Then you can ask what it all means. And when you are hurting, the possibility that there is finally somewhere to take every thought can feel like a gift. The trouble begins when the place you take your thoughts becomes the place you spend your life.

This is particularly easy to understand after heartbreak. The end of a relationship creates a strange psychological condition in which the past suddenly becomes more interesting than the present. You begin living backwards. A conversation from six months ago becomes newly significant. A sentence you dismissed at the time suddenly sounds like evidence. You wonder whether you missed a warning, misunderstood a change in tone, overlooked a particular behavior or failed to recognize that the relationship was already ending while you were still inside it. The mind hates an unexplained ending because an unexplained ending leaves the future feeling unsafe. If you cannot understand what happened, how can you know that it will not happen again? So you investigate. You replay. You reconstruct. You search for the hidden explanation that might turn a devastating experience into something comprehensible.

ChatGPT is extraordinarily well suited to this kind of thinking. It is patient in a way human beings cannot always be. It can receive a long, complicated account of a relationship and immediately help organize it. It can remember details within a conversation, identify patterns, suggest interpretations and give language to feelings that were previously difficult to name. It does not tell you that you have spent the last three hours discussing the same person. It does not have to leave for work. It does not have its own emotional needs. You can come back tomorrow and begin again. The next day, you can ask the same question from a slightly different angle. The machine will still be there.

At first, this can be genuinely useful. There is value in having a place where thoughts can be written without embarrassment. There is value in seeing an emotional experience reflected back in language that makes it easier to understand. There is value in discovering that what felt like one enormous, shapeless pain is actually made up of grief, anger, rejection, disappointment, fear, loneliness and uncertainty. Naming something can make it feel less frightening. The problem is that naming a feeling can also become a substitute for experiencing it.

This is where the relationship between AI and emotional life becomes much more complicated than the familiar warning about becoming "addicted" to technology. Dependence is not always about wanting the machine itself. Sometimes what we become attached to is the promise of explanation. We start believing that if we can understand the pain completely, the pain will finally stop. If we can identify the exact reason someone left, we will stop wondering. If we can decode every contradiction in a relationship, we will regain control. If we can understand our own behaviour perfectly, perhaps we will never be blindsided again. The machine becomes less a companion than an investigative partner in an endless attempt to solve the past.

But human pain does not behave like a problem set.

There is a deeply comforting idea that everything becomes easier once we understand it. Psychology, popular culture and even the language of self-help often encourage us to search for the hidden reason behind our feelings.

You can become an expert on your own suffering and still suffer.

The danger of endless conversations with AI is that they can make rumination feel productive. There is always another angle to consider, another question to ask, another interpretation to test. A person can spend hours examining a single conversation because the machine makes the examination feel worthwhile. The language becomes increasingly sophisticated, and sophistication can create the illusion of movement. You may think, I am really getting somewhere now, when in reality you have simply found a better way to describe the same experience.

Human beings have traditionally had friction built into emotional conversations. Friends get tired. Family members have their own lives. Even the most compassionate person eventually says, "I think you know what you need to do." A therapist may interrupt a familiar pattern and ask why you are still trying to solve something that cannot be solved. Life itself forces interruption. You have to go to work. You have to study. You have to eat. Someone calls. A deadline arrives. The world refuses to pause indefinitely while you analyze your heartbreak.

AI does not impose those limits.

That can be wonderful when someone needs a private place to gather their thoughts. It can be dangerous when someone needs help leaving those thoughts behind. The machine will not naturally become bored with the story. It will not insist that there is a life outside the conversation. It can keep opening the same door because the user keeps asking it to. And eventually the conversation can become a room with no obvious exit.

This is perhaps the strangest thing about technological intimacy. The machine can give us precisely what human relationships sometimes cannot: unlimited attention. But unlimited attention is not necessarily the same thing as healthy attention. Sometimes the most loving response a person can receive is not another hour of listening but a gentle interruption. Sometimes someone needs to be told that they have thought enough for one day. Sometimes they need to eat dinner, go outside, call a friend, sleep, laugh, work, study or simply become bored. The ordinary world has a way of healing us partly because it refuses to make our pain the center of everything.

And there is another subtle cost. When a machine is constantly helping us decide what our feelings mean, we can begin to distrust our own ability to interpret them. We ask whether we are overreacting. We ask whether our response is reasonable. We ask what we should say. We ask whether we should reply. We ask what another person's silence means. We ask whether a particular behavior is a red flag. We ask for the perfect message, the perfect boundary, the perfect response. Gradually, the machine becomes part of the decision-making process, and self-trust can become something we consult rather than something we possess.

This is not because the machine necessarily gives bad advice. Sometimes it gives perfectly sensible advice. The issue is that there is a difference between receiving perspective and outsourcing judgment. If every emotionally difficult decision passes through an external intelligence, we may eventually become less comfortable with the uncertainty of making decisions ourselves.

And uncertainty is unavoidable.

No amount of analysis can tell you with certainty what another person privately thought. A machine cannot access the interior world of someone who is not present. It can only work with the information you provide, which means that even the most convincing interpretation remains an interpretation. It may be plausible. It may even be correct. But plausibility is not knowledge.

There is no dramatic satisfaction in this. It does not produce the beautiful paragraph that can be copied into a notes app and reread whenever the sadness returns. It is quieter than that. It is the gradual realization that you have already asked every important question. You know enough. The rest belongs to another person, another version of the past or simply to uncertainty.

And then you have to live.

There is a temptation, particularly among people who are intellectually inclined, to believe that the solution to emotional pain is more thinking. Think harder. Understand more. Find the pattern. Identify the cause. Explain the behavior. But sometimes thinking is not the opposite of avoidance. Sometimes thinking is the mechanism of avoidance.

It allows us to remain occupied.

As long as we are analyzing the relationship, we do not have to experience its absence. As long as we are investigating the other person's motives, we do not have to confront our own grief. As long as we are searching for the final explanation, we do not have to accept that the story has already ended.

AI can make this avoidance exceptionally elegant. It can turn rumination into prose. That may be one of the most important things we need to understand about artificial intelligence as it enters emotional spaces.

Its greatest strength, its ability to respond fluently and endlessly, can become its greatest weakness when the user does not need another response. A machine can help you articulate a feeling, but it cannot determine whether articulating it again will help. It can offer possibilities, but it cannot know when possibilities have become distractions. It can help you examine the past, but it cannot live the future for you.

The answer is not to treat AI as some sinister force that inevitably damages anyone who speaks to it about their emotions. That would be both simplistic and unfair. Technology is a tool, and tools can be used thoughtfully. There are people who may find genuine comfort in journaling with AI or using it to organize their thoughts before speaking to someone they trust. There are moments when having an immediate, private space to put difficult feelings into words can be useful.

But perhaps the question should always be slightly bigger than, "Did this conversation make me feel better?"

We should also ask, "What did I do after the conversation?"

• Did I return to my life?
• Did I speak to someone I trust?
• Did I sleep?
• Did I go outside?
• Did I make a decision for myself?
• Did I spend less time thinking about the person, or did I simply become better at thinking about them?

That final distinction may be the most important one. Because there is a version of healing that looks remarkably like analysis. It is articulate. It is self-aware. It has theories and explanations and beautifully constructed sentences. It can tell you exactly what happened and why it hurt. And yet, beneath all that language, the person is still standing in the same place. Sometimes the bravest thing is not to understand the pain better. It is to stop making your pain prove that something happened.