Patients Increasingly Find AI-Produced Health Info Online – But Many Prefer Physicians
By Alisa Pierce

Almost half of Americans still consider physicians their most trusted source of medical information – yet a growing number find online health information produced by artificial intelligence (AI) at least somewhat reliable.  

As patients enter exam rooms with questions shaped by algorithms, physicians are working to provide clarity grounded in clinical expertise. 

A recent national survey from the Annenberg Public Policy Center of the University of Pennsylvania found nearly 8 in 10 U.S. adults are likely to search online for answers to specific questions about health symptoms or conditions – and nearly two-thirds of that group say they often come across AI-generated responses as top results on search engines like Google’s “AI Overview” or Bing’s “Copilot Answer.” 

According to the Annenberg survey, 63% of patients who saw AI summaries in their internet searches found the results at least somewhat reliable, while 31% said the answers “often” or “always” gave them the information they needed. Another 45% said AI “sometimes” provided the correct information. 

Nevertheless, 90% of those surveyed said they still trust physicians as trustworthy sources of health information and almost half (49%) still prefer clinicians’ education and experience over AI-generated recommendations.  

On the other hand, patients remain hesitant about physicians using AI tools in patient care – only 36% of the survey’s respondents said they were comfortable with physicians using AI in clinical settings. 

“The question of how physicians can talk to their patients about AI is usually answered by listening,” says League City family physician Renny Varghese, MD, a member of TMA’s Committee on Health Information Technology and Augmented Intelligence. “What are they scared of? What symptoms do they feel aren’t being properly addressed? From there, we’re able to reassure them about our use of AI, while underscoring its potential risks.”  

Dr. Varghese regularly uses AI transcription services within his practice, as do many of his colleagues. His patients use AI at home, too, and he’s regularly questioned both in-person and via online patient portals about AI-generated health information, with patients even “copy and pasting whatever ChatGPT generated into our practice’s online message box,” he said. 

Such instances, he says, serve as an opportunity for physicians to both listen to their patients’ concerns and communicate when and how AI should be used in the medical space. Namely, TMA’s policy on augmented intelligence in health care asserts the technology should be used as a tool set with physicians at the lead, rather than AI-based tools, which use data sets to make their decisions.  

Dr. Varghese’s approach has led his patients to become more comfortable with AI tools in appropriate settings – over the past two years, only three of his patients have declined when asked for consent to use AI tools during their visits. He also says open communication with patients makes them less likely to rely on AI-generated information rather than clinical expertise. 

Especially as AI tools base their summaries on numbers and talking points that have frequently been cited online, regardless of validity, their use can still cause “a lot of anxiety for patients,” says Philip Bernard, MD, chair of TMA’s Committee on HIT and Augmented Intelligence. 

That’s just one reason TMA works to ensure physicians sit securely at “the head of the [AI] table, and I think TMA is doing a wonderful job to enforce that concept,” Dr. Bernard said. 

His committee also continues to develop new ways to educate physicians on how to appropriately use AI-integrated technologies, and TMA policy promotes the safe and effective use of technology for practice efficiency, quality improvement, and population health while reducing physician burden.  

“I have no problem using AI for transcription purposes and use it every day. But it cannot replace the judgment of a physician … I would not trust ChatGPT to be my physician,” Dr. Bernard. “It’s important we understand the use cases as well as the limitations of this technology.” 

For more information, visit TMA’s comprehensive AI webpage.