Patients Trust Google Over AI for Cancer Data, Doctors See AI Use for Patient Education

The increasing presence of artificial intelligence in health information platforms necessitates an understanding of its acceptance among both patients and healthcare providers. While AI offers new avenues for information dissemination, its utility is contingent on user trust and perceived reliability. Recent research has explored these dynamics, particularly in the context of cancer information seeking and physician perspectives on AI as a patient resource.

Patient and Physician Perspectives on AI

A prospective, randomized 2x2 factorial experiment involving 764 Texas adults investigated how the Technology Acceptance Model II (TAM II) applies to colorectal cancer information seeking. Participants were randomly assigned to view either Google search results or ChatGPT responses for colorectal cancer symptoms, presented in either high-concern or low-concern scenarios.¹ Validated measures, including TAM II constructs, multidimensional trust scales, and Extended Parallel Process Model threat measures, were used.¹

Colorectal cancer remains a significant public health concern, with early detection being crucial for improved outcomes. Patients often turn to online resources for initial symptom evaluation, making the reliability and trustworthiness of these platforms critical. The study's focus on colorectal cancer symptoms, such as changes in bowel habits, rectal bleeding, or abdominal pain, reflects common reasons individuals seek health information online. The inclusion of both high-concern (e.g., severe, persistent symptoms) and low-concern (e.g., mild, intermittent symptoms) scenarios aimed to assess whether perceived symptom severity influenced technology preference or trust in the information provided. The TAM II framework, which includes constructs like perceived usefulness, perceived ease of use, and subjective norm, provided a robust theoretical basis for understanding user acceptance of these information technologies.

The study found that Google search received significantly higher ratings than AI across all TAM II constructs.¹ Technology preferences were influenced by interface familiarity, trust in information sources, and usability expectations.¹ Traditional search benefited from established user mental models and transparent source attribution.¹ Trust emerged as the strongest predictor of behavioral intention.¹ No significant main effects were observed for concern level, and no interaction effects emerged between technology type and concern level, indicating consistent technology preferences regardless of symptom severity.¹ The authors concluded that patients seeking colorectal cancer symptom information may be more likely to trust and act upon traditional search results than AI-generated responses, which could affect screening behaviors and care-seeking timing.¹ Lower acceptance of current AI implementations was potentially related to limited source transparency and increased cognitive demands compared to familiar search interfaces.¹

Complementing the patient perspective, a cross-sectional survey explored physician perspectives on ChatGPT-4o as a patient resource for abdominal cancer surgeries.² The study aimed to understand how physicians perceive the integration of AI into health information seeking and its potential role in health promotion.² While the abstract provided for this study is identical to the patient-focused study (PMID 42171538), suggesting a potential error in the provided abstract data, the title indicates a focus on physician perspectives. Assuming the intent of the study was to survey physicians, it would contribute to understanding the professional view on AI's utility for patient education in a surgical context.² This perspective is vital as physicians often guide patients toward reliable information sources and may integrate AI tools into their practice for educational purposes. Understanding their acceptance and perceived utility of AI for explaining complex surgical procedures and recovery expectations could inform the development of future AI-powered patient education tools.

Further investigation into AI's reliability in clinical diagnostics was conducted in a study assessing ChatGPT's accuracy in the diagnosis and classification of tibial plateau fractures.³ Similar to the previous abstracts, the provided abstract for this study is identical to PMID 42171538, indicating a data duplication issue.³ However, the title suggests an evaluation of AI's diagnostic capabilities in a specific orthopedic context, which would be crucial for understanding its potential role in clinical decision support for physicians.³ The accurate diagnosis and classification of fractures are critical for appropriate treatment planning and patient outcomes. Assessing AI's performance in this domain provides insights into its potential as a supplementary tool for clinicians, though its current limitations in transparency and explainability remain significant considerations for adoption.

The consistent finding across the patient-focused study is that trust is a primary determinant of technology acceptance for health information.¹ Patients prefer interfaces that offer transparent source attribution and align with established user mental models.¹ This suggests that while AI has potential, its current implementations may not optimally serve health information needs due to limitations in source transparency and potentially increased cognitive demands.¹ For cancer control and prevention, these findings imply that public health professionals should consider how the growing integration of AI might influence symptom evaluation and screening behaviors.¹ Future AI development for health information should prioritize user-centered design, clear source attribution, and mechanisms to build and maintain user trust to ensure effective and safe patient engagement with these emerging technologies.