Key takeaways
- In a random sample of 200 U.S. independent practices outside hospital systems, ChatGPT mentioned or cited 0 practices.
- The largest citation source was hospital staff directories (27.5%). Independent practices’ own websites accounted for about 23%, mostly from a small number of large groups.
- The gap narrowed for procedure-related questions, and independent practices accounted for most citations in dermatology and plastic surgery.
Study findings - How often do independent practices appear in ChatGPT answers?
This study was conducted by Halcy, an AI-services company serving independent practices. Its founder described the findings in a July 2026 article for the medical trade publication Medical Economics. It examined GPT-5.5 in May 2026.
- Random sample: 200 independent practices across 18 specialties and 20 cities; approximately 4,950 patient-style questions → 0 practices
- Citation sources (approximately 33,000): Hospital staff directories 27.5% · Independent practice websites about 23% · Specialty society and credential pages 22.1% · National physician directories 8.3%
- By specialty: Cardiology and orthopedics: hospital-affiliated physicians made up about 1 in 3, while independent practices were in the single digits. In dermatology and plastic surgery, most citations went to independent practices.
- By question type: The gap was widest for reputation and discovery questions, and narrowed only for procedure-related questions.
- Independent-practice mentions by city: Boston about 7% to Charlotte over 20%
- The contributor’s explanation: ChatGPT needs sources it can rely on before it can name a practice.
The Institute’s interpretation - Where can independent practices in Korea fit in?
This is a company’s own study, conducted by a business that sells services to independent practices, and its findings have not been independently verified. It is a correlational analysis of ChatGPT alone in the U.S.; Korean care settings and platforms may differ. The suggestions for independent practices are the Institute’s interpretation.
A domestic survey found that, in the past 3 months, 81.6% had used Naver and 54.5% had used ChatGPT for search (general search; Opensurvey). Another survey found that 48.9% of respondents had consulted generative AI about a health concern (Korea Press Foundation). We found no domestic data showing how independent practices appear in AI answers, so the U.S. findings should be read as a checklist, not an answer.
As the contributor explains, for a hospital’s name to appear in an answer, there must be a source describing that hospital. Independent practices should start by checking where those sources appear.
- Hospital staff directories are a source independent practices may lack. According to the contributor, physicians at large hospitals are automatically included in directories managed by their organizations. Independent practices may lack those sources unless they maintain their own physician profiles. We could not confirm whether the same structure applies at large hospitals in Korea. Still, independent practices do need to maintain their physician information themselves.
- Credential information is accessible regardless of practice size. The 22.1% figure for specialty-society and credential pages comes from the U.S. and cannot be applied directly to Korea. Even so, any hospital can check whether its external credential information is accurate.
- Procedure-related questions may be a better starting point than trying to cover everything at once. The gap was wide for reputation questions but narrowed for procedure questions. Start with clear, specific information about your main procedures. Procedure information must comply with the medical advertising standards in Korea’s Medical Service Act.
Where are the sources that describe our practice?
- Do physician profiles clearly list names, specialist qualifications, and main areas of care?
- Do external pages, such as those of medical societies or associations, show current physician information?
- Is there a separate page for each main procedure explaining who it is for, the process, and recovery?
Results for our specialty and local area matter more than national averages. Try entering several likely patient questions in Korean into ChatGPT and Naver, then record whether our practice appears and whether the information is accurate. That can help establish a baseline.
Sources
Nathan Woo (Halcy founder), “Why ChatGPT favors hospitals over independent practices,” Medical Economics (2026.7.20, company article)
Original: Medical Economics · Study: Halcy
Opensurvey, “AI 검색 트렌드 리포트 2026” (company announcement): Opensurvey Blog
Korea Press Foundation <미디어 서베이> 2026 Issue 1: Foundation materials · 기자협회보 · Yonhap News Agency (2026.4.15)