Four out of 10 GPs (38%) have had at least one patient express unrealistic expectations or dissatisfaction in a consultation in the last month after using AI to do their own research, a national Healthed poll of more than 1000 GPs has found.

Many GPs say AI has been helpful in terms of patients’ understanding of their health in general (55%) and their ability to effectively manage their conditions (45%).

But it can also come at a cost – 62% of GPs in the survey said AI use was having a negative impact on patient anxiety.

And nearly half (47%) felt patients’ use of AI was contributing to unreasonable pressure against GPs clinical judgement – while 43% noted a negative impact on the length of consultation time.

Patients bringing in AI generated health info is becoming the norm

A recent NewsGP poll of around 800 GPs found that 27% had patients bringing AI generated health information into consultations on a daily basis – while 30% said it happens weekly, 13% monthly, 19% rarely and 9% never.

Meanwhile, some patients appear to be using AI to ‘fact check’ their doctors, by providing it clinical details and asking if the recommendations or diagnosis they received were appropriate, AusDoc reported earlier this month, citing MDA national, who told the publication that patients were filing complaints based on the information from AI – often without the full clinical context.

A significant proportion of GPs in Healthed’s survey – 45% – said they felt pressure against their clinical judgement as a result of patients’ AI use.

“I feel challenged by information patients can get, almost like I have to ‘get it right,’ and agree with AI,” one doctor commented.

However, many GPs also noted improvements in self-management capabilities and engagement as a result of AI.

“Sometimes some weird ideas are generated, but I think they are more invested and more informed overall,” one GP commented.

“They become more knowledgeable and it is easier to discuss about health issues,” another GP said.

Many could see both pros and cons.

For example, using AI could help patients make lifestyle modifications that have a positive impact – but when it’s used to try to self-diagnose it tends to be “mostly negative, leading to long consultations and extensive testing on demand,” one GP said.

Another GP, who requests that patients bring their AI chat to discuss, said most people are reasonable, but “I feel they might be inclined to follow the app rather than my advice.”  A recent patient who had been using retatrutide, ‘Wolverine stack’ and testosterone for 5-6 months presented with symptoms including malaise. “He did not discuss this with me prior to buying them… I still could not convince him that the injections (especially since not TGA-approved) were dangerous and AI and social media had told him otherwise,” the GP recounted.

What your colleagues are saying

An opportunity for patient education

“I show the patients how I look things up online. We discuss AI answers as easy to find but I show them more reliable, pertinent medical websites for their problems and often print out some info for them to take away and read.”

“If you spend time in discussions, most patients are happy.”

“Currently, AI assists patients with preparing for their consultation with the doctor, but they still trust the doctor’s experience and management.”

A mixed bag

“Patients’ responses to AI use are very variable, however, the overwhelming majority realise that their research may have led them to a wrong conclusion. In general, they are keen to verify with the doctor the information they believe they have gleaned. At present this is mostly positive as it can streamline a consultation towards their main concerns and they may already have a framework of understanding that they did not previously have.

A small percentage of patients will have become very anxious and have a very skewed interpretation of their research. This will require extra time to sort out what assumptions (false or otherwise) are guiding their thought process. I believe this is a well observed phenomenon but thankfully it is not the most common outcome of AI guided research in my practice.”

Net negative

“It gets very annoying when patients use chatGPT and at the end of the consult say ‘Oh yes, that’s the treatment chatGPT said.’ It’s actually offensive.”

“Personally, it feels like the AI feeds them what they want to hear so had one even saying it affirmed her choice of a coffee enema for bowel cancer. No idea what she entered into the AI engine to get that sort of result.”