Andreessen: Doctor ChatGPT outperforms 99% of physicians. Is that true?

      Marc Andreessen has made a provocative assertion regarding health. The billionaire investor claims that "Doctor ChatGPT" is already superior to 99% of human doctors. This assertion has been met with strong opposition from medical professionals and lacks robust peer-reviewed support.

      Andreessen’s comment was made during an early June appearance on Joe Rogan's podcast. The New York Post highlighted his statement, and it gained traction after being shared by the prediction-market account Polymarket on X on June 29. The statistic is striking, but its foundation is weak.

      The Claim

      Andreessen did not provide any data to substantiate the 99% claim. He heads Andreessen Horowitz, one of the most influential venture capital firms in the tech industry, which lends weight to his words. He stated to Rogan, “99% of the time, the answer that I’m getting from the AI is better than I would get from talking to basically almost any expert.” The Post transformed this into the assertion about doctors.

      Many people are indeed trying this. OpenAI reports that over 40 million individuals inquire about their health using ChatGPT daily. Given such numbers, even a minor error rate could lead to millions of potentially dangerous responses, particularly for those who cannot afford a proper medical consultation.

      What Studies Actually Show

      Research presents a more nuanced view. A 2024 study published in the Journal of Medical Internet Research had both doctors and ChatGPT-4 answer 100 real patient inquiries. Patients perceived the chatbot as more empathetic and helpful, but specialists identified 15 of its responses as potentially harmful, as reported by IBTimes UK, cautioning that laypersons might struggle to distinguish between safe and dangerous information.

      A more recent study raises further concerns. In February 2026, a Nature Medicine article led by Dr. Ashwin Ramaswamy from Mount Sinai evaluated OpenAI’s ChatGPT Health across 60 clinical scenarios. The tool failed to recommend emergency care for 51.6% of genuine emergencies, instead suggesting routine appointments. One such case involved a patient experiencing respiratory failure.

      This aligns with a known trend. While AI has become proficient at identifying likely diagnoses, it struggles more with assessing risk and making treatment decisions, which are crucial aspects of actual medical practice.

      Following the Money

      Andreessen has financial interests linked to this assertion as well. His firm has invested in health AI startups, such as Hippocratic AI, Ambience Healthcare, and Abridge. Additionally, it manages a $500 million biotech fund backed by pharmaceutical giant Eli Lilly.

      This doesn’t necessarily render him incorrect, but it does indicate that his optimism is not impartial. When someone promoting Doctor ChatGPT stands to gain if people accept his view, the claim warrants careful scrutiny.

      The Prompt That Backfired

      There are also reasons to question his understanding of the tool. In May, Andreessen posted a lengthy “super prompt” on X instructing ChatGPT to “never hallucinate or make anything up” and to verify every fact.

      Engineers pointed out the clear flaw: one cannot eliminate hallucination merely through instructions. This phenomenon is inherent to the functioning of large language models. AI critic Gary Marcus found it “hilarious (and maybe a little bit scary)” that Andreessen had not grasped this concept.

      Why It Matters

      The implications are significant. Millions now rely on a fluent and confident chatbot, mistaking its fluency for medical reliability—two concepts that are not equivalent.

      Doctors do not claim that AI is useless; many are already utilizing it, and diagnostic tools are continually advancing. Dr. Adam Rodman, a Harvard hospitalist studying AI in medicine, advocates for a sensible approach: never use a chatbot to triage an emergency, and view it as a supplementary resource to a human consultation, not as a substitute.

      This is the point that Andreessen’s soundbite overlooks. The public is already cautious about AI enthusiasm, especially concerning health, a wariness that is justified. Doctor ChatGPT may indeed have its benefits, but based on current evidence, it is not better than 99% of doctors.

      This article addresses medical decisions and personal health. AI tools can be helpful starting points, but for any urgent or serious concerns, it is essential to consult a qualified healthcare professional.