Other Side Of Ai: Does It Aid Or Hinder Mental Health?

After being diagnosed in 2018 with attention-deficit/hyperactivity disorder (ADHD), which often comes with the baggage of emotional dysregulation, time blindness and task paralysis, Michael (not his real name) lost his job.

“My manager lost patience with me,” he said.

What happened next was a spiral of self-flagellation, anger over his disability costing a job he liked, and depression when comparing himself with his siblings and others his age, who are married with children. He was also lonely. 

Then, in 2022, OpenAI (an artificial intelligence research and deployment company) released ChatGPT, which quickly became popular. Other AI chatbots followed suit. And Michael soon found himself using AI chatbots for all aspects of his life.

He found AI helpful in managing his time and life, helping him structure and plan his work and play, and providing him with someone to talk to when lonely. He added that AI has also been great in helping him understand ADHD and traits associated with it, such as rejection sensitivity dysphoria (RSD), which is the inability to temper emotional reaction to a perceived or actual rejection, failure or criticism.

The difficulties of living with ADHD have also caused many people to develop mood disorders, including anxiety and depression.

“(I use AI in) so many ways. Even (for) family situations, loneliness, all these personal and professional (problems), I consult AI,” Michael told Bernama via Google Meet.

“I can even send AI screenshots of my messages for them to interpret (and) for me (to scan) for rejection. It's part of my RSD where I scan for rejection in people's responses a lot.” 

Michael is not alone in using AI to manage aspects of his mental health. Research has found that many use AI not just for productivity, but also for mental health. There are an estimated one billion AI users in the world.

‘Use of Generative AI for Mental Health Advice Among US Adolescents and Young Adults’, a report published in JAMA Network – a comprehensive consortium of peer-reviewed medical journals published by the American Medical Association – in November 2025, found that about 13 percent of 1,058 respondents in the United States admitted to using generative AI for mental health advice, with 22.2 percent of them aged 18 to 21.

In the United Kingdom, new polling by Censuswide at the behest of charity Mental Health UK found that one in three adults use AI chatbots for mental health assistance. Usage of AI chatbots peaks at 64 percent among the 25- to 34-year-olds, and 15 percent of the respondents aged 55 and above said they used AI chatbots for help. 

At the same time, reports of a mental health crisis have abounded, with news reports of suicides among youths and adults due to AI. Other than social media, mental health specialists and AI experts are concerned that the use of AI chatbots as companions is worsening the issue.

LONELINESS EPIDEMIC

Since ChatGPT and other chatbots like Gemini and Claude became commonplace in workplaces and on phones, research has shown mental health problems associated with heavy use of AI. 

According to the report ‘Enhancing Mental Health with Artificial Intelligence: Current Trends and Future Prospects’, published in the August 2024 volume of ‘Journal of Medicine, Surgery, and Public Health’, prominent mental ailments such as depression and anxiety cost the global economy about US$1 trillion annually in lost productivity.

Malaysia is not exempt from the loneliness epidemic. Using the Malaysia Global School‐based Student Health Survey 2012 – under the Ministry of Health’s National Health and Morbidity Survey 2012 – self‐administered questionnaire on 33,523 students, researchers found that the prevalence of loneliness among secondary school students in Malaysia stood at 16.2 percent, with females having the highest prevalence at 20.8 percent. (Students who responded “most of the time” or “always” to feeling lonely during the past 12 months were categorised as lonely.) 

While AI might seem the solution to the loneliness epidemic, a global health concern noted by the World Health Organisation since 2023, experts warn that it does not replace real people and interactions with other human beings.

On June 11, a Canadian mother Kristie Carrier filed a lawsuit against OpenAI, claiming that ChatGPT validated her daughter Alice’s thoughts on suicide, which finally led to her suicide.

Despite reports of negative effects of AI on mental health, the tech has expanded, with even some psychiatric and mental facilities testing out using AI chatbots to provide therapy.

Institute of Strategic and International Studies Malaysia Fellow Tan Heang Lee told Bernama that AI can be a good coach for social situations. 

”There was a study done in the UK and they found that some adolescents are, in fact, transferring the skills that they are practicing with the AI chatbots to real-world situations, so in those contexts, AI becomes a practice ground,” she said.

AI and tech enthusiasts, such as Meta chief executive officer Mark Zuckerberg, have said AI is a way to solve loneliness. However, health and AI experts say chatbots could worsen loneliness as more and more people depend on AI for emotional connection.

The problem occurs when the AI companions replace human companionship or human therapists.

DESIGNED DEPENDENCE

To understand the worry, one must first understand how AI chatbots work. Tan said many AI-dependent cases begin with work-related tasks and productivity assistance. As they continue to chat and divulge personal information and questions, many users form emotional connections with the chatbot.

“Oftentimes people are using these ‘companions’ to meet a kind of emotional need and these companions act as a friend, sometimes as a coach, a confidant, a therapist, and sometimes even as a lover,” she said.

“One of the concerns is that AI companion use can facilitate emotional attachment, dependence and also compulsive use.”

By design, AI chatbots are designed to prolong engagement, such as by asking follow-up questions. Tan said AI companies are incentivised to create emotional attachment and also dependence to increase engagement.

“I think you know that (follow-up questions) is the AI chatbot version of infinite scroll,” she said, adding that interactions with chatbots are often frictionless, incredibly supportive and less exhausting than dealing with humans.

She said while AI makes interaction easier, it should never replace human connection as doing so could stifle personal growth and learning social skills.

“But instead, you know, it's really incentivised to keep the user coming back for more and to create that attachment so that they keep using it, right?”

Clinical psychologist Izzat Zaid said AI must be considered as a personal assistant or an intern, a tool to be used rather than to replace a friend, family or partner. While AI may be good to assist the socially anxious to practice and navigate tricky social situations, she reminded that AI could never really replace human interaction.

She added that it definitely should not replace human therapy as therapists are bound by ethics, unlike AI chatbots.

“When (we) therapists do our work, what we want is for our clients to be empowered with skills so that one day they no longer need us,” Izzat said.

Tan agreed. 

”I think that's one of the contrasts if you were to compare an AI companion with a therapist. It’s also important to note that professional therapists are bound by professional ethics to encourage the independence of clients. So, the goal of therapy really would be to eventually wean off the therapist, but this is not necessarily the case for AI companions,” she said.

IMPOSING LIMITATIONS

Too much of a good thing is an adage for a reason. And like the phrase, no matter how beneficial you think AI has been or can be for your life and mental health, it is always best to impose limits on its usage.

Associate Prof Dr Aini Suzana Ariffin, who leads the initiatives for AI governance and regulation in Malaysia and is also a Fellow Expert in the Centre for Artificial Intelligence and Robotics at Universiti Teknologi Malaysia, said prompt engineering is one way to prevent dependence on AI chatbots.

“It depends on the prompt engineering. As long as your instruction entails what you want AI to do, it will help you,” she said.

Some instructions include requiring limits on the interactions and not allowing follow-up questions, or engaging personally with AI. Or it is just as simple as making sure you have a real-life person for you to check and verify the information AI chatbots give you.

“A friend (of mine) uses AI to validate how she feels what she’s going through, whether it makes sense or not. But she uses me as her (reality) check.

“She would call me ‘Jat GPT’,” Izzat said, laughing.

The keyword here is human connection.

Tan said there needs to be more green infrastructure, like parks and playgrounds, to encourage people to pursue recreational activities outside.

“We need more parks and playgrounds for children, and where people can actually gather and meet and connect. I think the fact that people are turning to chatbots to get companionship is really symptomatic of a lack of social connection in our society, in our daily lives,” she said.

As for Michael, he is not worried about becoming dependent on his AI chatbot as the benefits still outweigh the risks at the moment. He has safeguards in place to ensure he does not become dependent on AI, such as going out and hanging out with his friends.

And if all else fails, he has the best backup. 

“I put the phone down,” he said.