Cancer patients are asking ChatGPT: oncologist Sidiropoulou warns that without a medical history, AI misinforms

By Los Angeles Times (LA Times Studios) · November 3, 2026.

Zacharoula Sidiropoulou, a breast cancer surgeon and head of the Breast Unit at the Lisbon West Local Health Unit, spoke with JBMS at the ESMO 2026 congress (Europe's leading oncology meeting) about the growing influence of artificial intelligence on how patients understand cancer and make treatment decisions. The piece is a video interview; the only material available is its summary, so we stick to what is stated there.

According to that summary, Sidiropoulou explains that medical information from ChatGPT and other online sources can be misleading when it lacks the context of each patient's diagnosis and medical history. She also mentions the European Cancer Information Portals project, which aims to improve access to reliable information. In addition, she addresses the importance of discussing fertility preservation before treatment, the risks of cancer misinformation and the need for advances in breast cancer to reach all patients.

A note on transparency: the content is produced by LA Times Studios, which according to its own disclaimer is independent of the newspaper's newsroom and does not constitute medical advice. The summary contains no new figures or data; it is a clinician's testimony, not a study.

**Our take.** The warning is sensible and points to the real failure: a generic answer about breast cancer can be correct in the abstract and wrong for a specific patient, because stage, subtype, age or reproductive wishes change the decision. The risk is not that AI always "lies," but that it sounds confident while knowing nothing about the person asking.

The fertility example is telling. It is exactly the kind of topic a chatbot will not raise on its own initiative and that a clinician must bring up before treatment begins. That proactiveness, grounded in knowing the person, is what a generic assistant cannot replicate.

That is why we believe the answer is not to ban patients from consulting AI, which is unfeasible, but to offer better alternatives: public, verified portals, such as the European project the doctor cites, and tools that work with medical records within clinical pathways, under human supervision. In general, the value worth defending in health care lies in verification and traceability, not just in the fluency of the answer.

In the short term, this means more work for medical teams, who will have to correct preconceptions patients bring from the internet. In the long term, the same technology, well integrated with clinical data and validated, could bring more personalized and accessible oncology within reach, which is why Sidiropoulou's emphasis on equitable access matters: progress only counts if it reaches everyone.