Beyond the hype: A journalist’s guide to decoding AI in clinical care
By Jennifer L.W. Fink, Wisconsin Health Journalism Fellow
“Most media coverage of AI in health care swings between utopian promises and dystopian fears,” said LaVonne Roberts, a health, tech and science reporter, during her lightning talk at Health Journalism 2025 in Los Angeles. “We cannot be reporting on health care if we don’t understand AI.”
Artificial intelligence, she said, isn’t coming; it’s here. AI-enabled tools are already used in clinical practice, and billions of dollars are being funneled into developing health-related technologies, Roberts said. When covering this topic, she advised journalists to rely on basic reporting tools: critical thinking, curiosity, and concrete frameworks, instead of press releases.
How to tell if an AI story has legs
Roberts introduced the acronym PIE to help journalists evaluate AI tools:
- Problem – What real-world problem is the tool attempting to solve with AI? Whether it’s reducing hospital readmissions, shortening drug development timelines, or addressing workforce shortages, the problem must be real and urgent. AI tools that directly affect clinical care are of much more interest to the public than tools that address obscure or backend problems.
- Implementation – Look at feasibility and workflow integration. Will the tool save clinicians time? Make health care more human? Technologies that give doctors (and other health care providers) more time to spend with patients are more likely to succeed – and to be worth a story..
- Evidence – Does the tool have a robust evidence base? Roberts described a large language model AI tool created by Dr. Eric Oermann and other researchers at the NYU Grossman School of Medicine that can predict hospital readmission more effectively than human doctors can. The researchers reported their findings in a June 2023 Nature article.
When considering evidence, be sure to also ask about baseline comparisons: How was this problem or process typically handled without AI? What are the costs – financial and otherwise – of both options? Ask about failure as well; understanding limitations is key to identifying whether tech has real clinical potential or is just smoke and mirrors.
“If they can’t tell you how it fails, that’s a massive red flag,” Roberts said. “This is not a story that has legs, in my opinion.”
Jennifer L.W. Fink is an independent health, parenting, and education journalist. She is also a 2025 AHCJ Wisconsin Health Journalism fellow.
