The use of generative AI tools by cancer patients and caregivers: A cross-sectional survey of awareness, attitudes, and experiences.

Abstract

1612 Background: Generative artificial intelligence (AI) tools offer novel avenues for patient support, yet data on their adoption in oncology remain limited. This study evaluated the awareness, attitudes, and usage patterns of these tools among cancer patients and caregivers to inform clinical integration. Methods: A descriptive, cross-sectional survey was conducted at a tertiary oncology center. Adult cancer patients and caregivers completed an anonymous, structured questionnaire assessing sociodemographics, digital literacy, AI usage patterns, and perceptions. Multivariate logistic regression identified independent predictors of AI use. Results: A total of 400 participants (68% patients, 32% caregivers; mean age 59 years) were included. While 75% were aware of AI chatbots, only 24% used them for cancer-related information. Multivariate analysis revealed that university-level education (adjusted odds ratio [aOR] = 4.47, p < 0.001), younger age (≤55 years; aOR = 2.64, p < 0.001), and recent diagnosis (<1 year; aOR = 2.16, p = 0.003) were independent predictors of AI use (Table 1). Notably, 69.1% of users did not disclose their AI use to their physician. Despite this, no user reported decreased trust in their physician; in fact, 26.8% reported an increase. A substantial 73.3% of all participants expressed a willingness to use a physician-verified AI tool. Subgroup analysis revealed divergent usage patterns: caregivers used AI more regularly (weekly/daily: 54.3% vs. 30.6%; p = 0.022) and focused on researching medical terminology (62.9% vs. 38.7%; p = 0.022) and alternative medicine (34.3% vs. 16.1%; p = 0.040), whereas patients were significantly more likely to seek emotional support (22.6% vs. 5.7%; p = 0.032). Conclusions: Generative AI use among cancer patients and caregivers remains moderate and reflects a marked digital divide driven by age and education. The prevalence of “silent” AI use raises significant patient safety concerns related to the risks of unverified information. Nevertheless, the strong demand for physician-endorsed tools highlights a clear opportunity for clinical integration. Oncologists should proactively address AI usage to mitigate misinformation risks and ensure safe integration into clinical practice. Independent predictors of generative AI use for cancer information (n=400). Parameter aOR 95% CI p value Age (≤55 years vs. >55 years) 2.64 1.594 - 4.381 <0.001 Education level (University or higher vs. lower) 4.47 2.195 - 9.109 <0.001 Time since diagnosis (<1 year vs. ≥1 year) 2.16 1.289 - 3.625 0.003 Abbreviations: aOR, adjusted odds ratio; CI, confidence interval.