Comment on ‘Understandability and Actionability of Artificial Intelligence‐Assisted Lymphedema Education Material in Patients Undergoing Breast Cancer Surgery: Expert Evaluation’

We read with great interest the study by Maraş Baydoğan et al. (2025) examining the understandability and actionability of artificial intelligence–assisted lymphedema education material for patients undergoing breast cancer surgery. The investigation contributes to the expanding dialogue on artificial intelligence integration into clinical education, and the authors' methodological rigour offers a foundation on which future inquiry can build. Nevertheless, several dimensions of the study warrant further examination to strengthen the interpretability, clinical utility and translational relevance of the findings. The authors developed a methodological design based on expert evaluation of an artificial intelligence–generated educational booklet. This approach brings value by incorporating informed professional judgement, particularly in a domain where inaccurate guidance may influence complications such as breast cancer–related lymphedema. However, exclusive reliance on experts creates an interpretive framework that does not fully capture the informational realities of postoperative patients. Clinical communication often becomes strained in the early postoperative period when patients process risk information through layers of pain, fatigue and emotional burden. Expert assessments function as proxies that do not reflect how patients absorb, prioritise or operationalise materials in daily self-care (Pineda et al. 2025). A key challenge arises from the assumption that readability and expert-rated clarity can predict patient comprehension. Patients with differing health literacy levels interpret educational content in diverse ways, particularly for chronic conditions that require behavioural vigilance. Self-management of lymphedema depends on daily preventive actions, such as limb protection, early symptom recognition and adherence to activity modifications. Without observing how patients navigate the booklet's logic, instructions and visual structure, it is difficult to determine whether the content supports real-world decision-making. Furthermore, the methodological approach does not allow the exploration of contextual barriers competing priorities, cultural beliefs or misconceptions that often influence postoperative self-management behaviour. Even when content is linguistically accessible, its meaning may not translate into sustained practice unless it aligns with the lived experience. The study design also raises questions regarding the representativeness of the expert panel. Although the authors selected individuals with doctoral training in surgical nursing, the homogeneity of the evaluators may limit the breadth of perspectives relevant to patient education, such as occupational therapy, physiotherapy, psychology and survivorship care. These disciplines frequently address lymphedema prevention and could offer insights into stepwise behavioural reinforcement or the alignment of educational content with functional recovery trajectories. A multidisciplinary interpretive framework would have strengthened the evaluation and broadened the clinical grounding (Koçbilek et al. 2025). The reported intraclass correlation coefficient of 0.74 indicates moderate agreement among experts. However, a broad confidence interval introduces substantial uncertainty, with its lower bound crossing zero. In clinical research, such variability affects not only the statistical interpretation but also the reliability of educational material as a consistent guidance tool. When experts disagree on the clarity or utility of specific instructions, this suggests that patients may encounter similar ambiguities. Self-management guidance must offer operational clarity to reduce postoperative complications. Therefore, variability in expert assessment signals potential inconsistencies that could influence limb-protection decisions, symptom monitoring or escalation of concerns. The mean understandability score was high, yet its dispersion indicated that several items elicited different judgements. Notably, elements related to visual aids, captions, tables and procedural cues demonstrated lower agreement. These components directly influence patients' operationalization of preventive strategies. For example, unclear visual representations of swelling progression or improper depiction of limb positions during gentle exercise can mislead patients in their attempts to interpret their symptoms or apply preventive pressure gradients. Similarly, inconsistent actionability scoring points to uncertainty regarding whether the educational content provides sufficiently explicit steps for patients. Actionability is particularly relevant in lymphedema prevention, where success depends on the consistent application of behaviour sequences rather than general awareness (Herron 2025). The authors correctly calculated the descriptive measures; however, the clinical meaning of the observed standard deviation in actionability (18.47) deserves further consideration. This wide dispersion suggests that experts perceived the variable feasibility of recommended actions under real-world conditions. Some instructions may require equipment, physical ability or time commitments that do not align with postoperative limitations. Educational materials that overestimate patients' capacity to perform certain actions risk cognitive overload and disengagement. Refining the behavioural architecture of material clear sequencing, conditional guidance and concrete thresholds for escalation would enhance its clinical interpretability and reduce the likelihood of misapplied instructions. The authors used multiple readability formulas to evaluate text complexity, thereby providing a comprehensive analysis of linguistic accessibility. However, readability scores do not reflect conceptual clarity, contextual appropriateness or clinical accuracy. Artificial intelligence systems are known to generate fluent but imprecise clinical statements, and this study reported instances in which the material received lower agreement on visual and table-related items. Such discrepancies often indicate structural gaps in procedural communication. For conditions such as lymphedema, in which patients must differentiate normal postoperative swelling from early pathological changes, precise visual representation holds significant clinical value. The absence of detailed visual instructions reduces the material's ability to support symptom appraisal and anticipatory self-care. The translational context of artificial intelligence–generated content also requires scrutiny. The educational material was developed through a series of prompts; however, prompt engineering constraints may limit the completeness or nuances of technical explanations. Reproducibility and version integrity remain uncertain without transparent documentation of the prompting process. Artificial intelligence systems may produce similar content differently over time owing to model updates, meaning that educational materials generated at one point may not be replicable later without clear procedural guidance. This issue affects clinical governance, as educational content disseminated to patients requires stable and verifiable sources to prevent inadvertent misinformation. Furthermore, this study's reliance on Turkish readability indices narrows its interpretive scope to a single linguistic and cultural context. Healthcare systems are increasingly serving diverse populations, and lymphedema prevention relies on sustained self-management across cultural settings. Translational fidelity of how well educational materials maintain meaning when adapted to other languages, health beliefs or communication norms is essential for broader applicability. Without evaluating cross-cultural transferability, the generalizability of the findings remains limited. Clinical applicability depends on how well the educational material aligns with postoperative realities and patient capacities. The authors noted that the content included risk factors, preventive strategies, nutritional guidance, daily management techniques and exercise recommendations. Although comprehensiveness is beneficial, it may introduce a cognitive burden when patients attempt to navigate unfamiliar information during recovery. Lymphedema prevention requires prioritisation; some actions carry a high protective value, whereas others contribute marginally. Educational materials that present all information with equal weights can dilute critical messages and reduce adherence. Individualised strategies are central to the prevention of lymphedema. Patients experience diverse anatomical characteristics, treatment exposure, physical function limitations and lifestyle constraints. A general booklet cannot capture the variability in the extent of axillary lymph node dissection, adjuvant therapy effects or postoperative mobility restrictions. Without adaptive guidance, patients may overapply or underapply instructions, leading to unnecessary anxiety or insufficient preventive behaviours. Artificial intelligence offers the potential for dynamic, personalised education; however, the static booklet format limits this capability. Integrating conditional logic, symptom-triggered decision pathways and personalised goal suggestions could greatly enhance clinical relevance. Another issue relates to the sequencing of instructions. Several PEMAT items concerning actionable steps received lower agreement among experts. This suggests that the recommended behaviours may not have been broken down into sufficiently discrete tasks. Effective postoperative education typically relies on scaffolding and simple actions leading to more involved behaviours as recovery progresses. Instructional materials that lack temporal structure risk overwhelming patients, particularly when discussing exercise techniques, skincare routines and weight management strategies immediately after surgery. Finally, clinical applicability relies on the alignment of educational materials with patterns of professional practice. Healthcare providers who deliver postoperative counselling must be able to reference, endorse and integrate content into routine care. If artificial intelligence–generated materials contain phrasing, visuals or conceptual structures that diverge from common clinical language, they may inadvertently create discordance between professional advice and written instructions. The expert panel identified concerns related to visual and tabular clarity, which may affect providers' willingness to adopt the material in practice. The study by Baydoğan et al. advances the conversation on artificial intelligence–assisted patient education and provides early insights into the potential of such tools to support lymphedema prevention after breast cancer surgery. The methodological design offers valuable preliminary evidence regarding expert-rated understandability and actionability; however, several aspects warrant refinement to enhance its clinical meaning. Greater attention to real-world interpretive processes, behavioural feasibility, visual clarity, reproducibility of artificial intelligence content and individualised guidance would strengthen the translational utility of such educational materials. Despite these considerations, this study meaningfully contributes to the emerging evidence base and sets the stage for future investigations that integrate patient perspectives, cross-disciplinary expertise and adaptive artificial intelligence systems into postoperative education strategies. S.G. conceptualised the commentary, provided clinical interpretation and critically reviewed the manuscript. P.J. led the drafting, analytical synthesis and final editing of the manuscript, serving as the corresponding author. J.P. contributed to data interpretation, literature contextualization and revisions. J.S. provided overall supervision, methodological validation and final approval. All authors read and approved the final manuscript prior to submission. The authors have nothing to report. Generative AI tools (ChatGPT-5 and Paperpal) were used only for grammar and formatting refinement. All critical interpretation, analysis and argumentation were authored solely by the listed authors. The authors have nothing to report. The authors have nothing to report. The authors declare no conflicts of interest. The authors have nothing to report.