TLDR: New research indicates a growing trend where individuals struggle to differentiate between medical advice from artificial intelligence and human doctors, often placing more trust in AI, even when the information provided is inaccurate. This phenomenon is leading to significant challenges in healthcare, with patients arriving at appointments with AI-generated diagnoses and treatment plans, sometimes requiring doctors to spend valuable time correcting misinformation.
Recent studies from institutions across North America, including the University of British Columbia (UBC), the University of Waterloo, MIT Media Lab, and a survey by Pearl.com, highlight a concerning shift in public perception regarding medical advice: people are increasingly unable to distinguish between AI-generated responses and those from human doctors, and often place greater trust in AI.
A UBC study, published on October 12, 2025, found that large language models (LLMs) like ChatGPT are not only providing medical information but are often perceived as more convincing and empathetic than real people. Dr. Vered Shwartz, a UBC assistant professor of computer science, noted, ‘The conversations with large language models were more persuasive than the ones with people. They were also considered more empathetic.’ She added, ‘When it’s written in such an authoritative way and has answers to everything you ask it, of course you assume it knows what it’s talking about.’ This human-like tone and confident presentation can easily sway users, even when the information is inaccurate.
The implications for healthcare are already being felt. Family physicians, such as Dr. Cassandra Stiller-Moldovan in Colwood, B.C., report that patients are frequently arriving at appointments with pre-conceived diagnoses or treatment suggestions derived from AI tools. ‘Oftentimes people get it bang on,’ Dr. Stiller-Moldovan stated, ‘But other times, you spend a lot of time convincing them that that’s not exactly what’s going on. That education piece takes a lot of time.’ This diverts valuable time from actual patient care to correcting misinformation.
Accuracy remains a significant concern. Research from the University of Waterloo evaluated ChatGPT version 4 on open-ended medical questions and found that only 31 percent of its responses were accurate in diagnosing health problems, meaning it was wrong more often than it was right. Despite this, studies from MIT Media Lab and GAIA-GEN UNAM revealed that participants could not effectively distinguish between AI and doctor responses, and showed a preference for AI-generated advice, rating high-accuracy AI responses as significantly more valid, trustworthy, and complete. Alarmingly, even low-accuracy AI responses were perceived similarly to doctors’ responses, leading participants to indicate a high tendency to follow potentially harmful advice.
A survey by Pearl.com, cited in a Newsweek report, further underscores this trend, indicating that 22 percent of Americans have taken incorrect medical advice from AI. The survey also found that 23 percent of respondents trusted AI’s medical advice more than a doctor’s, with roughly 10 percent even trusting AI’s recommendation for a cancer treatment regime over a physician’s. Andy Kurtzig, founder and CEO of Pearl.com, highlighted the underlying societal issues, stating, ‘This report reveals a stark reality: millions of Americans are being locked out of essential professional help by cost and distance… While it’s valid to turn to AI for answers, our findings show it’s a dangerous gamble.’
Experts acknowledge AI’s advancements while cautioning against over-reliance. Anant Vinjamoori, MD, Chief Longevity Officer at Superpower, noted, ‘The first thing to recognize is that medical AI has gotten much better. There’s now strong evidence showing that these systems can answer many questions at the level of an experienced physician.’ However, he added, ‘General-purpose AI may sound confident, but it mostly regurgitates broad answers and struggles with the nuance of an individual’s context.’ Kumar Dharmarajan, co-founder and chief medical officer of World Class Health, emphasized that while ‘AI can instantly synthesize a vast trove of medical knowledge and often deliver sound, even superior, guidance 24/7—and it will only get better,’ it ‘lacks the nuanced judgment, ethical training, and real-world experience that are essential for high-stakes situations. An algorithm can inform, but only a human professional adds the required wisdom.’
Also Read:
- The Evolving Paradigm of Trust in Agentic AI: Opportunities, Challenges, and Human-AI Collaboration
- Generative AI Earns Significantly Higher Trust from IT and Business Leaders, New Report Reveals
The increased trust in potentially inaccurate AI-generated medical advice can lead to misdiagnosis, delayed or inappropriate treatment, and ultimately, harmful consequences for individuals. Researchers and medical professionals advocate for the careful integration of AI systems in collaboration with human experts to ensure accuracy and prevent the spread of misinformation, while still harnessing the benefits of this cutting-edge technology.
