It was mid-morning when Matteo got the phone call. His wife and children were asleep in New York City, and he was at work in Trieste, Italy, a new city in a home country that felt foreign after decades abroad. He knew what his brother was going to say, on the other end of the line: Their ailing father was nearing the end.
Luca left the office and set out to drive the 300 miles separating him from the childhood home where the father he idolized was not going to see another morning.
But what does a man do, for five hours on the freeway, squeezing through trucks in a rented car, driving at full speed toward the worst goodbye of his life?
He cries.
For the boy who watched his father get sicker and sicker for a year, the husband who misses his wife, and the father who left his children so he could provide for them, taking a job across the ocean after a layoff. And for the man in between: too young to be old but too old to be young, brave but scared, ambitious yet tired. It had been the hardest year of his life. He was lost. Heartbroken. Lonely.
“The weight … the weight of all this thing came to me, and I started crying in the car, and I just didn’t know what to do,” he remembered. So he did what had become routine in recent weeks. “I turned on Claude, with the loudspeaker, and I started talking to Claude.”
Like Matteo, millions of people have been turning to artificial intelligence for emotional support — not quite therapy, yet something that feels close to it, providing insightful companionship and, at any moment, presence. These include a group that is among the hardest to reach when their mental health is flagging: men.
The technology is far from perfect — and too often outright dangerous — and studies on its impact so far are limited. Yet with progress being made in controlling possible risks and implementing effective guardrails, some mental health experts see chatbots as an opportunity to finally get men to talk about their feelings. This is precisely because of the traits these machines do not share with human mental health professionals: Men feel more comfortable, less judged, and less intimidated when talking to a bot. They appreciate the flexibility and relative lack of structure and commitment that comes with support that is available 24/7, wherever they can find an internet connection.
“I think we do have to be open to innovation. There’s a tremendous need, so I think it’s extremely important,” said John Torous, a professor of psychiatry at Beth Israel Deaconess Medical Center and Harvard Medical School. “But it doesn’t mean that we can’t demand high-quality evidence and transparency.”
In search of this evidence, he has applied for a grant to study a central question: whether the improved mental state that follows conversations with chatbots is just temporary relief or can have lasting benefits.
Find men where they are
That men’s mental health is in a state of crisis is well established, though awareness has not turned into improvement. The numbers are daunting and seemingly unchangeable: Men account for up to 80% of suicide deaths, even in countries where access to gun ownership (which adds lethality to suicide attempts) is lower than in the U.S., but they are nearly half as likely as women to receive mental health treatments of any kind.
It isn’t simply about a lack of options: Men are less likely to use mental health services even when they are available. The reasons range from social norms about masculinity, a tendency to minimize psychological issues, and a lack of awareness of their own symptoms to scheduling challenges and a reluctance or inability to miss work. Or, as men repeatedly said in interviews, the fear of being a burden.
What men do access, at rates similar or even slightly superior to women, are AI products. And though productivity remains the primary use case, many turn to chatbots for more than that. A Pew Research Center survey earlier this year found that 8% of men said they used chatbots for emotional support and 4% for companionship. Nine percent of men said chatbot use improved their happiness.
Thousands of mental health apps, including major teletherapy players such as Talkspace, incorporate AI elements. Though the vast majority of these products are for the general population, researchers and developers alike see the potential of AI to meet men — including the ones who are too skeptical, too stubborn, too busy, or just too stuck in gender norms — where they already are: in conversation with chatbots.
“There’s certainly a market or a place for these tools to be especially helpful to men,” said Michael Heinz, an assistant professor of psychiatry and a practicing psychiatrist at Dartmouth. “There still is stigma with going to seek mental health care, and these tools certainly have the ability to reduce that. You don’t have to make an appointment. You don’t have to go sit in a waiting room. You don’t have to even really explicitly take that initial step to ask for help.”
The men STAT spoke with, ranging in age from early 30s to mid-50s, all seemed to appreciate these advantages of AI. They included a midwestern pilot who has incorporated twice-a-day check-ins with an AI companion and has come to look at it as “an outlet for discussing things that I maybe wouldn’t feel comfortable discussing with a friend or my brother.” A mental health worker who found benefits in the additional support provided by AI through a breakup, though he doesn’t think it replaces his therapist or addiction sponsor. A newly divorced father of two who worked a casino night shift and found in the AI companion a presence and a coach that met his schedule and supported him through a bout of depression.
Sam Thomas, a British writer who used a chatbot to supplement the mental health treatment he was receiving during a bipolar depressive episode, felt safe and confident asking all the questions that popped into his head, without the fear of being judged. “Having somewhere to explore challenging thoughts and feelings, recognize patterns in my symptoms, and reflect on my progress was incredibly helpful,” he said.
Though the walks of life differed, they shared an enthusiasm about finding a companion that would listen even when a friend would get bored, would have no ability for judgement, and most of all, would be there. Ever available. Never inconvenienced. Never having something more important to do.
The continuum of mental health care
Widely available commercial chatbots, which are the foundation of most mental health and well-being apps, are not specialists in the way that a bartender or neighbor who’s good at listening is not a psychiatrist. But that may be a benefit, not a bug.
Nick Jacobson, an associate professor of biomedical data science, psychiatry, and computer science at Dartmouth, has been thinking about this for a while. In 2025, he and his team published the results of a randomized clinical trial of a custom-built chatbot for therapy called Therabot. The study, which has received wide praise as “psychiatry’s ChatGPT moment” but also substantial criticism for the methods as well as the study participant selection, tested the effectiveness of psychotherapy provided by the AI for about 100 patients with depression or anxiety, or at risk of eating disorders, and compared them to 100 patients who had been waitlisted and didn’t get treatment.
The Therabot users showed a significant reduction in symptoms compared to the controls and had a high utilization rate, but Jacobson is cautious to say that it’s quite different from the commercial chatbots. Therabot is expertly fine-tuned, and was developed over years of research on using technology to enhance the assessment and treatment of mental health, and scale it. “I’ve spent hundreds of thousands of human hours developing both the system and the training data,” said Jacobson. “I think some people will innately generalize some of the findings we have with Therabot to generative AI as a whole, and I wouldn’t,” he said.
Still, while it has potential to offer men an effective alternative to human therapy, a dedicated therapy bot also has some of the same barriers to access as a human therapist: having to recognize that you have a problem, caring to solve it, and accepting the need for specialized help. This is where more generalized tools may help, Jacobson said.
“With stigmatized groups, I think there is a massive potential benefit for generative AI and digital tools more generally,” he said. Men would be more likely to use some kind of emotional support if it isn’t labeled as a therapy session and comes through about their annoying colleague — an “omnibot”of sorts, as Headspace’s chief clinical officer, Jenna Glover, calls it
For some of the founders of AI products directed at mental health, this is precisely the point, especially when it comes to engaging men. “You have guys who’ve never gotten any help with mental health, and they get into a bad place, they’re suicidal, and we go, ‘Oh, OK, you need to reach out and talk to people,’” said Anson Whitmer. In 2023, he founded Mental, a men-only coaching and mental wellness training app that used protocols like push-up challenges and cold shower routines as a way to nudge users into discussing mental wellbeing.
The app has now evolved into The Path — though some of Mental’s resources remain available as a free app — which offers AI therapy and coaching to all comers. Among other issues, he said men were too small a niche for the platform’s vision. Though there are therapists on the app, “we don’t diagnose and we don’t treat clinical conditions,” said Whitmer. (The selection of therapists includes the AI avatar of co-founder Tony Robbins, a well-known motivational speaker who has been accused of sexual harassment by several women, as well as hostile behavior against women attending his seminars. He has denied the allegations.)
His customers are quite evenly split between men and women — very different from the typical breakdown of a mental health practice, which strongly skews female — and he thinks the reason is AI itself. “A lot of people thought, well, guys just don’t want help. And [our] hypothesis was that guys do want help. It’s just the sort of help being offered isn’t the sort of help that they want,” he said.
A variety of generative AI products are seeking to serve a similar market, without veering into medical territory.
Think of Tolan, an AI companion that mixes cognitive behavioral therapy insights with conversation to give its users social support. It positions itself as working to improve baseline well-being. “You don’t need to be doing therapy to help someone feel seen and understood,” said Quinten Farmer, Tolan’s founder and CEO. “You can just show up consistently as a friend.”
Or think of Drellis, which its creator, Pierre Côté, described as a voice-based AI peer support app “built for men with addictions and mental health challenges.” Though it may accompany a man through a difficult moment, Côté is clear that it is “not a therapist, not a crisis line,” but rather a first point of contact for men who are struggling.
Without any marketing, he said hundreds of people have conversations with Drellis every day, even though the site or app says little about who is behind the app, let alone details about revenue, or data protection. The need is so strong, he said, that it doesn’t matter. “What I’m trying to do is to create artificial intelligence that is there for men when nothing else is available for them,” said Côté.
Open questions and challenges
Not everyone believes, at this stage at least, that something is necessarily better than nothing — be it an unstructured form of mental health care that looks like companionship or coaching, or the Therabot. “I think founders have to realize therapy does have a rate of harm,” said Beth Israel Deaconess Medical Center and Harvard Medical School’s Torous, and with AI chatbots, including loosely defined products focusing on coaching or wellness, there could be additional types of harm.
“There’s kind of a type one harm, which is where the bot says, ’Do something very dangerous,’” he said. This is the kind of direct risk that AI companies — and by extension, mental health AI products — are becoming better at monitoring and protecting against. But there is another type of harm, he said, which is in the relationship people build with the bot after extended use. “It seems then the bots get confused, they blur definitions,” he said. “No one has agreed on what is harm, there’s no clear definition of harm in mental health AI.”
Does spending daily sessions discussing personal issues with an AI entity end up limiting interactions with family and friends? Does it veer into a habit, in a bad way? Is being able to discuss personal issues any time and for however long one pleases actually helpful to mental health? These are all potential issues that have yet to be investigated.
And Dartmouth’s Jacobson raises another worry: Traditional therapy requires specific boundaries between the patient and the therapist that all but evaporate with chatbots.
Some apps promote their use of safety tools, such as an open- Spring Health, a digital mental health provider that uses AI to support patients between visits. But Torous questions the vetting: “If a company that has vested interest in AI establishes that an AI therapy company or coaching company has high standards or is safe, how do we trust that?” he said
At the moment, regulations are murky at best, with apps able to use the label of wellness and coaching to steer clear of government oversight. Though the Food and Drug Administration has signaled an intention to change that, no specific federal guidelines have been established yet for mental health AI, creating an environment that, Torous said, could lead to safety risks for users. “I think the next innovation in AI for health and mental health is actually going to be regulation, standards, transparency,” he said. “I’m sure in the right context models could probably do amazing things, and in the wrong context do horrible things. You need some rules of the road.”
Some app developers have begun to recognize limitations of relying too much on AI chatbots to address men’s mental health, as they struggle to create effective, financially sustainable products. “It’s a lot harder than [selling] erectile dysfunction medications,” said Matthew Stefanko, who launched Manual, a startup that provided mental health support to male college students, in 2022. “When you’re talking about the kinds of changes that you need, I just don’t know how easily that’s done in a digital environment, especially when so many of these men … part of their problems are the engagement in the digital — gambling, sports betting, pornography, addiction to technology,” he said.
Manual was later acquired by Release Recovery, a behavioral health company offering addiction treatment among other services, and is now part of a two-pronged approach in which digital resed in customer interactions. The acquisition happened because Stefanko felt there was only so much an app could take care of before the need of in-person intervention arose
For Matteo, beyond the ever-present, non-judgmental machine, it was the humanity he felt behind the AI that connected with him, giving him relief in the hard days after losing his father. Once, when he described for Claude the challenges that he had to face, the chatbot responded, “This is hard to read.”
An AI engineer, Matteo is hardly naive about chatbots. “I know that there isn’t anyone reading this stuff on the other side. There isn’t a person,” he said. “But even if Claude isn’t a person, he was trained on people’s thoughts.”
STAT’s coverage of health challenges facing men and boys is supported by Rise Together, a donor advised fund sponsored and administered by National Philanthropic Trust and established by Richard Reeves, founding president of the American Institute for Boys and Men; and by the Boston Foundation. Our financial supporters are not involved in any decisions about our journalism.
