A report has raised concerns that people are using AI chatbots for mental health support instead of regulated services.
The research, commissioned by mental health charity Mind, found that three in five people who had used chatbots for their mental health in the past year had done so instead of seeking formal support.
The findings also point to an increase in safeguarding concerns involving AI-related harm, including misleading advice and the risk of users becoming dependent on chatbots.
Rosie Weatherley, information content manager at Mind, said: “Many of us are using AI for lots of things, including to support our mental health.
“Whether that’s for emotional support and validation, easily digestible information, or something which appears to be listening and responding in a timely way. This is often because we can’t get what we need elsewhere.
“But what we’re finding – and what concerns us – is there are just as many harms as there are uses of AI tools.
“These can affect everyone, not just people experiencing mental health problems. Evidence shows us that even casual use of generative AI can lead to harm, as AI can be misleading or fuel dependency.
Among respondents who had used chatbots instead of services such as their GP, NHS talking therapies or counselling, 31 per cent said they preferred AI.
A further 18 per cent said they could not access formal support, while 11 per cent reported that existing options did not meet their needs.
Talking therapies are structured treatments delivered by trained professionals to help people understand and manage thoughts, feelings and behaviours.
Mind said it had received no safeguarding reports involving AI before 2026, but such cases accounted for nearly 10 per cent of all concerns reported to the charity since January.
Safeguarding involves protecting people from abuse, exploitation or other forms of harm.
The charity warned that while chatbots may provide immediate information or emotional support, they do not necessarily offer the protections and oversight found in regulated services.
Lily Shervington, a 26-year-old mindfulness teacher from Worcestershire who experiences obsessive-compulsive disorder, raised concerns about the use of AI for mental health advice.
Obsessive-compulsive disorder, known as OCD, can involve unwanted and distressing thoughts alongside repetitive behaviours or mental routines intended to reduce anxiety.
She said: “I often get caught up in believing things that aren’t real.
“My OCD involves a lot of ‘magical thinking’ so I believe I can control a lot more than I can, like keeping loved ones alive. I’m consumed by intrusive thoughts.
“AI, to me, represents a distorted version of reality. It’s ‘human like’ without being human and that scares me.
“I understand that it’s not trying to be a human, however, people treat it like one, and I believe that has dangerous consequences.
“Another issue from an OCD perspective is the constant question-asking in return. This feeds into the cyclical cycle of OCD and traps the individual in further thought loops. I think it’s detrimental to people with OCD.”
Despite the growing use of chatbots, 84 per cent of respondents said access to advice or support from a person alongside AI tools was important.
The charity said this suggested that users did not always regard the technology as a direct replacement for human care.
It is calling for a same-day, open-access model for in-person mental health support.
Nearly four in five respondents, or 79 per cent, also said stronger regulation of AI chatbots was important.
Mind has launched information accredited by the Patient Information Forum to help people assess whether they should use AI for mental health support.
The resource includes guidance on recognising when chatbot use may be causing harm and managing mental health while using AI tools.
The charity also launched its Mental Health in a World of AI commission on 9 July.
The 12-month initiative will gather evidence about potential harms, inequalities and ways the technology could be used more safely.
Its priorities include improving access to support for underserved communities, producing clearer guidance and campaigning for stronger accountability for technology companies.
Weatherley said: “While so much is unknown, we are particularly concerned about the long-term impacts of using AI tools.
“For example, we’re aware of how AI chatbots can offer short term support to reduce loneliness but at the same time, how this can lead to people becoming more isolated over time.
“We want to make sure that if people choose to get support through AI, that they are receiving accurate and responsible information.”
