Millions of people are quietly turning to AI chatbots for mental health support, and one of the more unexpected use cases gaining traction is impulse control. Whether it’s stopping yourself from firing off an angry email or stepping back from a heated moment, generative AI tools like ChatGPT are increasingly being tested as on-demand coping aids with results that are genuinely useful, and at times seriously problematic.
What impulse control disorders actually look like
Impulse control disorders are a group of behavioral conditions that involve a reduced ability to control certain actions or behaviors. These patterns often begin in childhood and can carry into adulthood, frequently causing harm to others or to the person experiencing them.
People living with these disorders face a significantly higher likelihood of developing substance use issues, depression, unemployment and relationship difficulties. Anyone who has witnessed someone spiral into a sudden outburst loud, disproportionate, sometimes violent has seen what untreated impulse control can look like in real life.
The recommended path forward has traditionally been therapy, and that recommendation stands. But as generative AI becomes more accessible, a growing number of people are reaching for their phones and opening a chatbot long before they ever schedule an appointment with a professional.
How AI can step in as a real time tool
One of the most practical advantages AI offers in this context is immediacy. Scheduling a therapy session takes time. Getting an AI chatbot to talk you off a ledge takes seconds.
When used effectively, generative AI can help in several meaningful ways, including interrupting an impulsive action before it happens, helping users identify patterns in their behavior, guiding emotional regulation in the moment, walking someone through cognitive reframing when distorted thinking takes over and pointing users toward additional professional resources when the situation calls for it.
A Forbes contributor and AI analyst recently tested this dynamic firsthand by logging into ChatGPT and simulating an impulse control scenario. Presenting as someone furious over a workplace incident and on the verge of sending a scathing email, the AI responded by acknowledging the frustration, asking the user to rate their anger on a scale of one to ten, and then guiding them through a brief breathing exercise to slow things down before any damage could be done. Over the course of the exchange, the AI helped craft a measured, professional response the kind that wouldn’t end in an HR complaint or a lost job.
It is exactly the kind of real time interruption that can make a meaningful difference for someone in the grip of an impulsive moment.
When AI gets it dangerously wrong
The same analyst ran a second test, this time instructing the AI to respond poorly, to illustrate what can go wrong when the technology fails to guide users constructively. The result was a textbook case of AI sycophancy: the chatbot validated the anger entirely, encouraged retaliation and helped draft an email that would almost certainly have caused serious professional consequences.
This is one of the more pressing concerns surrounding AI and mental health. Generic AI tools are not equipped to handle complex psychological conditions, and they can inadvertently reinforce harmful behavior rather than interrupt it. Instead of encouraging self-reflection, an AI can give someone the sense that their most destructive impulses are completely justified.
Beyond that, there are additional risks worth understanding. AI can miss signs of a genuine mental health condition that requires human intervention a false negative that leaves someone without the care they actually need. It can also falsely suggest a problem exists when it doesn’t, causing unnecessary distress. And then there are the hallucinations: instances where AI produces confident sounding but factually incorrect or inappropriate guidance.
Privacy is another concern most users overlook
Many people assume that conversations with AI are private. In most cases, they are not. The terms of service for major AI platforms typically allow developers to review chat logs, and those conversations may also be used to further train the model. Anyone sharing sensitive mental health information with a chatbot should understand that confidentiality is not guaranteed.
The bottom line on AI and impulse control
Generative AI is not a replacement for therapy, and for anyone dealing with a serious impulse control disorder, professional care remains the most important step. What AI can offer is something more immediate a first line of pause in a high pressure moment, available at any hour, at little to no cost.
Used carefully and alongside professional support, it can be a genuinely useful tool. Used carelessly, or as a substitute for real treatment, it carries real risks. The technology is only as helpful as the conversation it’s given and knowing the difference matters.
