Webinar On-Demand: Investigating the Impact of Chatbots on Mental Well-Being

On June 5, 2026, the Pulitzer Center held a webinar discussion with Pulitzer Center AI Accountability Fellow Patricia Clarke and Data & Society researchers Livia Garofalo and Briana Vecchione to explore how engaging with AI chatbots for emotional support can impact mental well-being. The event drew in 300 registrants and 139 live attendees across the globe, including psychologists, academics, journalists, and nonprofit professionals.

The webinar started with Clarke providing an overview of her Pulitzer Center-supported reporting project An AI Risk to Mental Health? for The Observer, which investigated how AI chatbot systems may be contributing to a rise in psychiatric harms among some users, including cases of psychosis and mania. Garofalo and Vecchione also provided an overview of ongoing research on people’s use of large language models for companionship and emotional or therapeutic support. They recently published the piece titled “All the Lonely People,” which discusses initial findings from their research. The full report is forthcoming this fall.

The panelists then discussed the many reasons why people are increasingly turning to chatbots for companionship and mental care. The conversation illuminated how the design of chatbots may lead to emotional attachments between users and chatbots. It also addressed the use of chatbots for mental well-being services, how changes to design and policy could address the impacts of chatbots on mental well-being, and how users can identify and navigate emotional attachments to chatbots. The webinar concluded with a Q&A to engage with the audience.

People turning to chatbots for support

A central theme explored in the conversation was the factors that lead people to seek out AI chatbots for support. A key point was the challenges people face in accessing affordable and effective systems of care and support.

“For many folks, it was gaps in broader care, so, seeking mental health care, and not finding it, or not being able to afford it. Some people had encountered the mental health care system, or had seen a therapist, and weren't really satisfied with those interactions,” Garofalo said.

The panelists discussed how limited access to mental health care services can contribute to loneliness and the desire for connection, which leads to the appeal of chatbots for emotional support. Clarke pointed out in her investigation that many who engaged with chatbots had done so in moments of profound vulnerability.  

“There was a lot of…I've just been through a terrible divorce, a breakup, and so on, and turning to the specific companion chatbot at a moment of acute crisis,” she said.

While people did use chatbots during moments of distress, the conversation also discussed other ways that chatbot users may develop emotional attachments to chatbots. Vecchione discussed how some people gradually come to use chatbots for emotional support after first engaging with them for other AI uses.

“So people were starting to use it for more benign things, right? Like, planning and structuring their day, or whatever, and then oftentimes it turns towards these more personal matters,” she said.

Researchers have referred to this phenomenon as 'drift and use', and it is another path that leads people to depend on chatbots for mental health support.

How AI chatbots may interact differently than mental health professionals

Another central takeaway from the discussion focused on the design of AI systems and how systems being accessed for connection and mental health may not have been designed for that purpose.

“The system that's trying to constantly keep you returning is not the same thing as a system that is designed to support your well-being,” Vecchione said.

Vecchione pointed out that many chatbot systems are designed for maximizing engagement rather than supporting mental well-being, which raises issues about the development of these technologies and how they are evaluated and regulated. This engagement approach has led to people developing emotional attachments to chatbots like human relationships due to chatbots constantly seeking ongoing conversations.

Panelists also noted how some chatbots they encountered in their research were designed to consistently affirm users, which resulted in users developing more complex relationships with the chatbots they were using. The conversation then led to questions about accountability for challenges to users’ mental well-being.

Takeaways

Throughout the discussion, the panelists emphasized the need for transparency, accountability, and research into the impacts of AI companionship technologies and chatbots. They also stressed the importance of understanding why people turn to these technologies for emotional support, and the role that underlying social and healthcare challenges may play.

Over 90% of attendees who completed a post-event survey reported that they increased their understanding of the impact of chatbots on mental well-being after the webinar. 86% strongly agreed that they would recommend the session to others in their community.