Cancer Patients Quietly Turn to AI Chatbots, and Most Keep It Secret From Their Doctors

Generative artificial intelligence has slipped quietly into the waiting rooms of oncology. A new study of 400 cancer patients and their caregivers at a tertiary oncology center reveals that while three-quarters of these individuals know about AI tools, only about a quarter actually use them for cancer-related information—and, strikingly, roughly seven in ten of those users never tell their oncologists. The findings, published in Supportive Care in Cancer, offer one of the most detailed portraits yet of how people living with cancer and the family members who care for them actually interact with chatbots and other generative AI systems, and the picture is one of widespread awareness, uneven adoption, and a pervasive culture of secrecy.

The research team, led by Rashad Ismayilov and colleagues at Başkent University Faculty of Medicine in Ankara, Turkey, conducted a descriptive, cross-sectional survey of adult patients and their informal caregivers. Participants completed anonymous, structured questionnaires covering sociodemographic characteristics, digital literacy, patterns of AI usage, and perceptions of the technology. Of the 400 participants, 68 percent were patients with a mean age of 61 years, while 32 percent were caregivers averaging 54 years of age. This design allowed the investigators to measure not just whether people had heard of tools such as ChatGPT, but who was actually turning to them for help navigating a cancer diagnosis, and why.

The headline numbers tell a story of a gap between curiosity and action. Seventy-five percent of participants were aware of generative AI tools, yet only 24 percent had used them to seek cancer-related information. That disparity is itself informative: awareness of AI is now nearly universal, but for many patients and caregivers the technology remains something they know about rather than something they use, particularly when the stakes involve a life-threatening illness. The authors describe overall usage as moderate, suggesting that AI has not yet become a routine companion in the cancer journey even as its profile has soared in the public consciousness.

Perhaps the most technically revealing part of the study lies in its multivariate logistic regression analysis, which identified independent predictors of AI use. University-level education was by far the strongest, with an adjusted odds ratio of 4.47, meaning highly educated participants were more than four times as likely to use AI tools. Younger age, defined as 55 years or below, carried an adjusted odds ratio of 2.64, and a recent diagnosis of less than one year was associated with an adjusted odds ratio of 2.16. All three predictors reached statistical significance. In plain terms, the people most likely to consult AI chatbots about cancer are young, well-educated, and newly diagnosed—a combination that points to a digital divide in how patients seek answers at the moment they may be most desperate for information.

That divide matters because it suggests the benefits of generative AI in oncology are not being distributed evenly. Patients with less formal education and older patients—groups that often face the greatest barriers to navigating complex health systems—are the least likely to harness a technology that could, in principle, help them understand their diagnosis, prepare questions for their physicians, or make sense of treatment options. The study’s authors argue that this pattern constitutes a distinct digital divide based on education and age, one that health systems will need to address deliberately if AI is to serve all patients rather than a digitally fluent subset.

The study also uncovered a striking divergence between patients and caregivers. Caregivers used AI tools more regularly than patients did, and the motivations behind that usage differed significantly between the two groups. This makes intuitive sense: caregivers often shoulder the burden of researching treatments, managing appointments, and translating dense medical information, and they may have more time, energy, and digital fluency than patients themselves, who are frequently exhausted by treatment. Any clinical strategy for integrating AI into supportive cancer care, the findings imply, should treat caregivers as a distinct user population with their own needs and behaviors rather than as proxies for patients.

The secrecy surrounding AI use may be the study’s most consequential finding. Sixty-nine percent of users did not disclose to their physician that they had consulted AI tools about their cancer. This silent usage creates a hidden layer of information-seeking that clinicians cannot see, verify, or correct. Because generative AI models can produce plausible-sounding but inaccurate medical information—a phenomenon documented in prior studies evaluating chatbot responses to cancer questions—unverified AI output circulating in patients’ minds without clinician oversight poses a genuine misinformation risk. The anonymity of the questionnaire may itself hint at why patients stay quiet: they may fear judgment, worry their doctor will disapprove, or simply not think of chatbot queries as something a physician would care about.

Yet the study also contains an unexpectedly reassuring message about trust. No AI users reported decreased trust in their oncologists as a result of consulting the technology. Even more striking, 73 percent of all participants said they would be willing to use AI tools if those tools were verified by their physician. That single statistic reframes the entire problem: patients and caregivers are not turning to chatbots out of distrust of their doctors, nor are they ideologically committed to replacing clinical expertise with algorithms. They are simply hungry for accessible information, and they would enthusiastically welcome physician-endorsed AI resources. The demand for verification is, in effect, an invitation for oncologists to engage rather than a signal of disengagement.

The authors conclude that oncologists must proactively invite open, non-judgmental discussions about AI usage in clinical encounters, so that they can clinically verify patients’ findings, mitigate misinformation risks, and ensure the safe integration of these tools into practice. In practical terms, this could mean asking patients directly whether they have consulted AI, recommending reputable AI-based resources, or explaining the limitations of chatbot-generated medical advice in a way that respects patients’ autonomy. Given that three-quarters of participants are already open to physician-verified AI, a clinician who normalizes the conversation stands to gain insight into what their patients are actually reading—and to correct errors before they harden into beliefs.

As generative AI continues its rapid spread through everyday life, this study suggests that cancer care is at a crossroads. Awareness is nearly universal, adoption is moderate and skewed toward the young and educated, and most usage happens in the shadows of the clinical relationship. But the strong appetite for physician-verified tools indicates that the divide between AI and oncology is not a chasm of mistrust—it is a gap in communication. Closing it may require nothing more technologically sophisticated than a question at the next appointment: have you been talking to a chatbot about your cancer?

Subject of Research: Awareness, attitudes, and usage of generative AI tools among cancer patients and their caregivers

Article Title: The use of generative AI tools among individuals with cancer and their caregivers: a cross-sectional study of awareness, attitudes, and experiences

Article References: Ismayilov, R., Oguz, A., Zenger, G. Y., Boluk, S. M., Farzaliyeva, A., Ramazanoglu, M. N., Akcali, Z., & Altundag, O. (2026). The use of generative AI tools among individuals with cancer and their caregivers: a cross-sectional study of awareness, attitudes, and experiences. Supportive Care in Cancer, 34(10), Article 935. https://doi.org/10.1007/s00520-026-11051-9

Image Credits: AI Generated

DOI: 10.1007/s00520-026-11051-9

Keywords: generative AI, cancer patients, caregivers, oncology, digital divide, ChatGPT, patient-physician communication, health misinformation, large language models, digital health, supportive care, cross-sectional study