Large Language Models as Health Information Tools: Patient Use and Trust in Otolaryngology
Beverly J. Fu, Itzel Rubio‐Jimenez, Alexander Ellman, Samridhi Semwal, Reza Vaezeafshar, Noel F. AyoubABSTRACT
Objectives
Patients increasingly use large language models (LLMs) for health information, yet their use patterns, perceptions, and impact on clinical communication in otolaryngology remain poorly characterized.
Methods
We conducted a cross‐sectional survey of adult patients at an academic otolaryngology clinic. The survey assessed demographics, LLM use for general or health‐related questions, and perceptions of LLM‐generated information, including usefulness, trust, and communication with clinicians. Primary outcomes were LLM use for health information overall and prior to the clinic visit. Secondary outcomes included perceived helpfulness, concordance with clinician information, preparedness, disclosure, and concerns.
Results
Among 59 respondents (mean age 50.1 years; 53.4% female), 71.2% reported prior LLM use, and 28.6% used an LLM to learn about their condition before the visit. All participants reported LLM information as helpful (50.0% very, 50.0% somewhat), that LLM information mostly (62.5%) or partially (25.0%) aligned with clinician explanations. Most felt more prepared for the visit (46.7% much more, 40.0% somewhat more), yet trust remained moderate: 32.7% reported mostly or complete distrust. While concerns about accuracy (82.7%) and privacy (50.0%) were common, 46.2% were likely to use LLMs again. Most users (93.8%) did not disclose LLM use, although 56.2% believed clinicians should be aware. Exploratory analyzes showed no significant differences in LLM use or perceptions by education, income, or social vulnerability index.
Conclusions
LLM use for health information is common among otolaryngology patients and associated with perceived benefit despite moderate trust and concerns. Engagement may be driven by factors beyond accuracy, including accessibility and perceived empathy.
Level of Evidence
3.