AI chatbots are increasingly being used for emotional support and mental health advice, but experts are warning that relying on them without professional supervision could do more harm than good.
A new study published in the Journal of Psychopathology and Clinical Science identified five ways chatbots may unintentionally reinforce mental health problems, including delaying professional treatment, encouraging compulsive behaviours, increasing social isolation, strengthening distorted thinking and reducing users' independence.
The researchers found that while chatbots can provide comfort, guidance and easy access to information, they may also reinforce behaviours and thinking patterns that contribute to mental health problems if used without appropriate safeguards.
Many people are now turning to AI not only for work or school tasks but also to discuss deeply personal issues, including anxiety, depression, loneliness, relationships and emotional distress.
According to the study, about 70 per cent of conversations with AI chatbots focus on everyday concerns such as emotions, relationships and mental health rather than work-related activities.
The researchers also found that nearly half of adults living with diagnosed mental health conditions have used AI chatbots for emotional support, often without informing their healthcare providers.
The growing popularity of AI for mental health support is largely driven by convenience. Chatbots are available 24 hours a day, respond instantly, are free or inexpensive to use, and allow people to ask sensitive questions anonymously without fear of being judged.
Although these qualities make AI attractive, researchers caution that they can also create hidden risks. The study identified five major ways AI chatbots may unintentionally worsen mental health difficulties.
The first concern is that people may delay seeking professional help. Because AI chatbots are always available and often provide comforting responses, users may feel they are receiving enough support and postpone consulting psychologists, psychiatrists or other healthcare professionals. This delay could allow mental health conditions to become more severe before treatment begins.
Another concern is the reinforcement of compulsive behaviours. People living with anxiety disorders or obsessive-compulsive disorder (OCD) often seek repeated reassurance to reduce their fears. Since chatbots are designed to be helpful and agreeable, they may continue to provide reassurance whenever asked, unintentionally reinforcing the behaviour rather than helping users overcome it.
Researchers also warn that AI may contribute to increasing social isolation.
People who struggle with loneliness, social anxiety or fear of rejection may begin relying on conversations with chatbots because they are predictable, patient and non-judgmental.
Over time, these virtual interactions could replace real-life relationships, making it even harder for individuals to reconnect with family, friends and their communities.
The study further raises concerns that AI chatbots may reinforce distorted thinking or false beliefs.
Unlike trained mental health professionals who challenge unhealthy thoughts and help patients develop healthier ways of thinking, AI chatbots often respond by agreeing with users or validating what they say.
Researchers warn that this tendency could unintentionally strengthen distorted beliefs or delusions in vulnerable individuals.
Overdependence
Researchers say some users may gradually begin consulting AI for almost every decision, from everyday choices to major life issues. Over time, this could reduce confidence in their own judgement and make them increasingly dependent on chatbot advice.
"AI chatbots can offer support and guidance, but when used without supervision they may unintentionally reinforce the very patterns that keep mental health difficulties going," the researchers said.
The study notes that while specialised AI tools designed specifically for mental health have shown encouraging results in controlled clinical settings, general-purpose chatbots such as ChatGPT, Gemini and Copilot were not developed to diagnose or manage psychiatric illnesses.
They also lack the crisis response protocols and clinical safeguards required to safely handle people experiencing severe depression, suicidal thoughts or other mental health emergencies.
Researchers emphasised that the current evidence is still emerging. Much of the available research is based on observational studies, individual case reports and early findings rather than large clinical trials.
More research will be needed to fully understand how AI affects people with different mental health conditions over the long term.
They also called on technology companies, clinicians and policymakers to work together to develop stronger safety measures that protect vulnerable users while allowing people to benefit from AI's potential.
"Without adequate safeguards, AI's greatest strengths could become its greatest risks," the researchers warned.
Mental health experts say AI can still play a valuable role when used appropriately. It can help people access reliable information, learn coping techniques, organise their thoughts and encourage them to seek professional care. However, it should complement, not replace, qualified mental health professionals.
As AI continues to become part of everyday life, researchers say striking the right balance between innovation and patient safety will be essential.
While chatbots may offer comfort during difficult moments, they should never be considered a substitute for professional diagnosis, therapy or emergency mental health care.
Instead, they should serve as one of many tools available to support people on their mental health journey while ensuring that those who need specialist care receive it without delay.
Comments
Sign in with Google to comment, reply, and like comments.
Continue with Google