ChatGPT is now a competitor to Doctor Google, according to a new report from Elsevier, an academic publishing company that also develops AI tools for medical professionals like Scopus AI and Reaxys. The findings, released today, come from a survey conducted earlier this year involving 2,206 doctors and nurses across 109 countries. Participants included 268 clinicians from North America, 1,170 from the Asia Pacific, 439 from Europe, 164 from Latin America, and 147 from the Middle East and Africa; 18 chose not to share their location.

The survey asked clinicians about their views on AI’s current role in healthcare and its potential future impact. Elsevier distributed the survey via email to healthcare workers who had previously published work or contributed to third-party panels, acknowledging that because the sample was not random, the results may not be universally applicable.

One of the primary concerns expressed by respondents was how patients are using ChatGPT and similar AI tools. Many arrive at appointments with incorrect assumptions based on what they’ve read online. This issue stems from the fact that these models often produce inaccurate information—OpenAI’s GPT-4o and GPT-4o-mini, for instance, reportedly generate false responses 30–50% of the time, according to internal testing.

This misinformation adds pressure on already overburdened healthcare professionals. In North America, 34% of time-constrained clinicians reported dealing with patients who ask many questions, while globally the figure was around 22%.

Jan Herzhoff, president of Elsevier’s global healthcare businesses and study sponsor, told Forbes that a more alarming trend could be patients opting out of visiting hospitals altogether and relying solely on AI or websites for health advice. He noted that more than half of U.S. clinicians expect most patients to self-diagnose instead of consulting a doctor within three years, though data on current trends remains unclear.

Despite these concerns, more clinicians are turning to AI themselves. The percentage of healthcare workers using AI in clinical settings has nearly doubled in the past year—from 26% to 48%. While many remain hopeful about AI’s potential to improve efficiency, few believe their institutions are effectively leveraging it to solve real-world issues.

Most surveyed clinicians anticipate that AI will help them save time, deliver faster and more accurate diagnoses, and ultimately enhance patient care within the next three years. Startups like K Health and Innovaccer, which have raised $384 million and $675 million respectively, are among those building such tools. According to PwC’s Strategy& team, the global AI healthcare market is projected to hit $868 billion by 2030.

Herzhoff emphasized that AI should act as a support system for clinicians, not a replacement. He sees particular promise in using AI for administrative tasks. Currently, clinicians use AI to detect harmful drug interactions before prescribing and to draft letters to patients. Among those who use AI, 50% regularly or always rely on general-purpose tools like ChatGPT or Gemini, compared to just 22% who frequently use specialized medical tools.

Part of the reason for this preference may stem from limited institutional support. Only one-third of respondents said their workplace offers adequate access to digital tools and AI training, while just 29% reported effective AI governance at their institution.

Ensuring that AI systems are trained on high-quality, peer-reviewed content is another top priority for clinicians—70% of doctors and 60% of nurses consider this essential.

Still, adoption faces challenges. Herzhoff acknowledged that although AI has the potential to save time in the long run, learning how to use it effectively—especially mastering prompt engineering—requires effort that many clinicians don’t have the bandwidth for right now. Nearly half (47%) of clinicians said fatigue has affected their ability to provide quality patient care.

“Paradoxically, they don’t have the time, and they don’t find the time to use these tools,” Herzhoff admitted.

While clinicians look forward to using AI to streamline daily tasks, they remain cautious about allowing it to influence clinical decisions. Only 16% currently use AI for decision-making, and 37% say they are unwilling to do so.

As one respondent put it: “AI should give me the information I need to make good decisions. But I won’t hand over responsibility for clinical judgment to an algorithm—I must retain control over final outcomes.”

The above is the detailed content of Doctors Think AI Shows Promise—But Worry It Will Mislead Patients. For more information, please follow other related articles on the PHP Chinese website!