Experts caution that relying on ChatGPT for medical guidance can expose one in eight Americans to misleading information. This warning comes as a new report from OpenAI, the creators of ChatGPT, reveals that 40 million Americans daily seek medical advice through the AI service, inquiring about symptoms and treatment options.

Furthermore, the report indicates that one-quarter of Americans consult ChatGPT weekly for health-related questions. Globally, healthcare topics account for one in every 20 messages sent to the platform.

Additionally, each week, users pose up to 2 million questions concerning health insurance, including inquiries about comparing plans, managing claims, and understanding billing processes.

OpenAI highlights that individuals in rural regions, where healthcare access is scarce, are more inclined to turn to ChatGPT, with 600,000 messages originating from these areas weekly.

Moreover, approximately 70% of health-related interactions occur outside regular clinic hours, underscoring a demand for medical advice during evenings and weekends.

And about seven in 10 health-related chat messages are sent outside of normal clinic hours, highlighting a need for health guidance in the evenings and on weekends. 

The report also found two in three American doctors have used ChatGPT in at least one case, while nearly half of nurses use AI weekly.  

Doctors who were not involved in the report told the Daily Mail that while ChatGPT has become more advanced and can help break down complex medical topics, it should not be a substitute for regular care.

A new report found 40 million Americans use ChatGPT every day for health information and advice (stock image)

Dr Anil Shah, a facial plastic surgeon at Shah Facial Plastics in Chicago, told the Daily Mail: ‘Used responsibly, AI has the potential to support patient education, improve visualization, and create more informed consultations. 

‘The problem is, we’re just not there yet.’

The findings come as OpenAI faces multiple lawsuits from people who claim they or their loved ones were harmed after using the technology. 

In California, 19-year-old college student Sam Nelson died of an overdose after asking ChatGPT for advice on taking drugs, his mother claims. 

SF Gate reported that, based on chat logs it reviewed, the service would first give a formal response insisting it could not help, but then if Nelson asked certain questions or phrased prompts a certain way, the tool was manipulated into providing answers. 

In another case, in April 2025, 16-year-old Adam Raine used ChatGPT to explore methods of ending his life, including what materials would be best for creating a noose. He later died by suicide.

Raine’s parents are involved in an ongoing lawsuit and seek ‘both damages for their son’s death and injunctive relief to prevent anything like this from ever happening again.’

The new report, published by OpenAI this week, found that three in five Americans view the healthcare system as ‘broken’ based on high costs, quality of care and a lack of nurses and other vital staff. 

Doctors warned that while ChatGPT can help break down complex medical topics, it is not a substitute for real medical care (stock image)

Dr Kathering Eisenberg, pictured above, said that while AI can make complex medical terms more accessible, she would ‘use it as a brainstorming tool’ instead of relying solely on it

In a ranking of the share of healthcare messages from hospital deserts, or areas that are at least 30 minutes from a hospital, Wyoming had the most messages at four percent, followed by Oregon and Montana with three percent.  

The team cited a survey of 1,042 adults who used the AI survey tool Knit in December 2025. They found 55 percent of those adults use AI to check or explore their symptoms, while 52 percent use it to ask questions at any time of day.

Additionally, 48 percent turned to ChatGPT to understand medical terms or instructions, while 44 percent used it to learn about treatment options. 

The OpenAI team cited several case studies in the report. Ayrin Santoso, from San Francisco, told the company she used ChatGPT to help coordinate care for her mother in Indonesia after she suffered sudden vision loss. 

And Dr Margie Albers, a family physician in rural Montana, said she uses Oracle Clinical Assist, which relies on OpenAI models, to help take notes and save time on clerical work.  

Samantha Marxen, a licensed clinical alcohol and drug counselor and clinical director at Cliffside Recovery in New Jersey, told the Daily Mail: ‘One of the services that ChatGPT can provide is to make medical language clearer that is sometimes difficult to decipher or even overwhelming.’

Dr Melissa Perry, Dean of George Mason University’s College of Public Health who has studied AI usage in the medical setting, told the Daily Mail: ‘When used appropriately, AI can improve health literacy and support more informed conversations with clinicians.’

Ayrin Santoso (left), from San Francisco, told the company she used ChatGPT to help coordinate care for her mother (right) in Indonesia after she suffered sudden vision loss

Dr Margie Albers (pictured here), a family physician in rural Montana, said she uses Oracle Clinical Assist, which relies on OpenAI models, to help take notes and save time on clerical work

However, Marxen warned that the ‘main problem is misdiagnosis.’ 

‘The AI could give generic information that does not fit one’s particular case, thus leading the person to misjudging the severity of the symptom,’ she said. 

On the flip side, AI could make a user think they are experiencing ‘the worst-case scenario,’ Marxen said. 

Dr Katherine Eisenberg, senior medical director of Dyna AI, told the Daily Mail: ‘I do think ChatGPT opens up access to more types of medical information to patients, but people have to understand that ChatGPT is trying to serve every need for every user and is not specifically built or optimized for medicine. 

‘I would suggest treating it as a brainstorming tool that is not a definitive opinion.’

She added: ‘I actually don’t think people need to stop using ChatGPT for medical information entirely. I think we’re beyond that point, with the amount of use it’s getting today.’

Instead, she suggested double checking all information from a reliable source like an academic center and avoid putting in sensitive personal information. ‘Most importantly, patients should feel comfortable telling their care team where information came from, so it can be discussed openly and put in context,’ she said.