The Promise and Perils of AI in Healthcare

Artificial intelligence is reshaping how we interact with health information — and how clinicians interact with us. Used wisely, it can support clearer communication and boost health literacy. But there are risks, and AI is no substitute for human connection, clinical expertise, or your own critical thinking.

We asked Tracy Mehan, MA, director of research translation and communication at Nationwide Children’s Hospital, about how AI can make a positive impact on our own healthcare and what caveats we should consider.

AI in the Exam Room

Benefit: Some providers use ambient listening software to transcribe medical appointments in real time, enabling doctors to focus more on their patients, explains Mehan. “It lets them step away from the screen so they can really hear you and interact with you, because they know the conversation is being recorded.”

Be aware: Ask if your clinician uses this kind of tool and, if so, how your privacy is protected. In this rapidly changing environment, HIPAA compliance and data security may vary, she cautions. You have the right to say no to AI use in the exam room.

AI in the Patient Portal

Benefit: Some portal platforms use AI to help generate after-visit summaries or replies to patient messages, enabling providers to respond to more inquiries more quickly. Mehan explains that any AI-crafted message in your patient portal should have been reviewed by your clinician before being posted. (Check with your healthcare provider if you’re unsure about their policy.)

Be aware: Electronic health records (EHR) systems that utilize AI may not be developed with health literacy in mind, and their responses may feel technical or unclear. If a message in your patient portal doesn’t make sense, follow up to get an answer you understand.

AI as Medical Translator

Benefit: Generative AI tools, like ChatGPT, Claude, and Perplexity, can be powerful allies if you use them skillfully, says Mehan. They can summarize medical information into plain language. “You can even ask these tools to create learning modules to teach you about a diagnosis or generate questions to bring to your next appointment,” she adds.

Be aware: These tools are superpowered predictive-text machines, she explains. They can’t read your mind, and they are limited by the questions you ask — as well as by how accurately they were trained. So don’t ask one question and take the response at face value. “Ask it to ask you questions so it can clarify your needs and give you better information,” Mehan suggests. Then fact-check what you find with a medical expert.

AI as Research Assistant (Not as Expert)

Benefit: AI can go deeper into medical literature more quickly than humans can. “I’ve heard stories of patients who’ve struggled for years with some sort of health condition, who put their symptoms into ChatGPT and all of a sudden have a potential diagnosis that hadn’t been considered,” Mehan notes.

Be aware: While AI-generated possibilities can feel empowering, they can also be biased by how their large language models were trained. AI can also hallucinate — inventing research and results that don’t exist; and with federal research funding slashed, AI results may contain outdated information. (This is why you should never stop at the “AI Overview” displayed on a web search, Mehan cautions.) Use AI to spark discussion with your healthcare provider rather than to self-diagnose.