More and more people are using generative artificial intelligence (AI) to support everyday tasks and work. ChatGPT is among the best-known and most accessible generative AI tools, offering personalised, seemingly credible responses to almost any question at no cost.
Generative AI tools have huge potential to help people understand health information. However, the responses are not reliably accurate. Depending on ChatGPT alone for health guidance can be unsafe and may also lead to needless worry.
Because generative AI is still a comparatively new and fast-evolving technology, evidence about how people use it is still emerging. Our new study reports the first Australian data on who is using ChatGPT to answer health questions and what they are using it for.
These findings can guide people on how to use this technology in health contexts, and what new capabilities are required to use it safely - in other words, to develop "AI health literacy".
Who uses ChatGPT for health? What do they ask?
In June 2024, we surveyed a nationally representative sample of more than 2,000 Australians about whether they had used ChatGPT to answer health questions.
One in ten (9.9%) said they had asked ChatGPT a health question during the first half of 2024.
Overall, respondents said they "somewhat" trusted ChatGPT, with an average rating of 3.1 out of 5.
We also found that using ChatGPT for health was more common among people with low health literacy, those born in a non-English-speaking country, and those who spoke a language other than English at home.
This pattern indicates ChatGPT may be filling a gap for people who struggle to engage with more traditional health information sources in Australia.
The most frequent types of health questions put to ChatGPT were about:
- learning about a health condition (48%)
- working out what symptoms might mean (37%)
- asking what actions to take (36%)
- understanding medical terminology (35%).
More than half (61%) had asked at least one question that would typically need clinical advice. We categorised these as "riskier". Asking ChatGPT what your symptoms might mean may provide a broad steer, but it cannot replace clinical advice.
People born in a non-English-speaking country, or those who spoke another language at home, were more likely to ask these kinds of questions.
Why does this matter?
Use of generative AI for health information is likely to increase. In our study, 39% of participants who had not yet used ChatGPT for health said they would consider doing so in the next six months.
If we include other tools - such as Google Gemini, Microsoft Copilot, and Meta AI - the overall share of people using generative AI for health information is likely to be higher still.
Importantly, our findings suggest people from culturally and linguistically diverse communities may be especially likely to use ChatGPT when looking for health information.
If people are also using ChatGPT to translate health information, that introduces additional complexity. Generative AI tools tend to be less accurate in languages other than English.
This means we need investment in services (human or machine) so that speaking another language is not a barrier to accessing high-quality health information.
What does 'AI health literacy' look like?
Generative AI is not going away, and it brings both benefits and risks for anyone using it to find health information.
On the positive side, the technology can be appealing to people who already face substantial barriers to accessing health care and health information. A major advantage is that it can provide instant health information written in a way that is easier to understand.
A recent review of studies found that generative AI tools are increasingly able to answer general health questions in plain language, although they were less accurate when questions involved more complex health topics.
This is particularly valuable given that much health information is written at a level that is too difficult for many people to understand - including during the pandemic.
On the other hand, people are using general-purpose AI tools as a source of health advice. This becomes more hazardous when the question requires clinical judgement and a broader view of the individual patient.
Case studies have already highlighted the potential harms of relying on general-purpose AI tools to decide whether someone should go to hospital.
Where else can you go for this information?
We need to support people to think carefully about the types of questions they are putting to AI tools, and to link them to appropriate services that can handle these "riskier" questions.
Organisations such as HealthDirect offer a free national helpline where you can speak to a registered nurse about whether you should go to hospital or see a doctor. HealthDirect also provides an online SymptomChecker tool to help you work out what to do next.
Although many Australian health agencies are developing AI policies, these largely focus on how health services and staff should use this technology.
There is an urgent need to build AI health literacy skills across the community. That need will increase as more people use AI tools for health, and it will also shift as the tools themselves change.
Julie Ayre, Postdoctoral Research Fellow, Sydney Health Literacy Lab, University of Sydney; Erin Cvejic, Associate Professor of Biostatistics, University of Sydney, and Kirsten McCaffery, NHMRC Principal Research Fellow, University of Sydney
This article is republished from The Conversation under a Creative Commons licence. Read the original article.
Comments
No comments yet. Be the first to comment!
Leave a Comment