Electronic health records, packed with crucial medical details, often resemble a foreign language to patients. Could AI bridge this communication gap? A new study explored using ChatGPT to translate complex multidisciplinary team (MDT) reports for colorectal and prostate cancer patients into plain English. These reports, dense with medical jargon and clinical assumptions, served as an ideal testing ground for ChatGPT’s ability to simplify medical information. Clinicians and non-patient lay people evaluated ChatGPT’s explanations, and focus groups—including patients, caregivers, and medical professionals—discussed the results. The study revealed several key challenges. Inaccuracies in interpreting medical abbreviations, providing incorrect URLs or test results, and using overly complex language were common. ChatGPT also struggled to personalize information, often providing generic advice and even potentially distressing statements. Furthermore, a lack of trust in AI-generated medical advice emerged as a significant barrier. Integrating such a system into clinical workflows, ensuring data privacy and security, and maintaining accuracy remain significant hurdles. While ChatGPT demonstrated potential for summarizing medical information in a patient-friendly manner, the consensus was clear: it’s not ready for clinical use. However, the study’s authors suggest that with further refinement—prompt engineering, fine-tuning, and rigorous safety checks—LLMs could eventually play a valuable role in drafting patient letters, which clinicians could then review and edit. The future of AI in healthcare hinges on addressing these key challenges. Further research focusing on patient and clinician needs, data security, and personalized communication will be crucial. Moreover, striking the right balance between AI assistance and human oversight will be essential to building trust and ensuring responsible implementation.