The primary research question addressed in this scoping review is: “What is the role of artificial intelligence in educating patients with breast cancer?”

A comprehensive literature search was conducted in 3 electronic databases: PubMed, ProQuest, and Scopus. All articles included were peer-reviewed publications from reputable international journals, published within the last 5 years (January 2020 to July 2025). Inclusion and exclusion criteria were determined using the Population, Concept, Context (PCC) model (Table 1).

Because each database uses different indexing systems and field codes, the search syntax was adapted accordingly. While the Boolean logic and core concepts remained consistent across databases—covering artificial intelligence, machine learning, chatbots, large language models, natural language processing, patient education, and breast cancer—the formatting was adjusted to optimize retrieval in each platform (Table 2).

This adaptation ensured conceptual equivalence while accounting for database-specific indexing features, such as the use of Medical Subject Headings (MeSH) in PubMed, abstract field tags (AB) in ProQuest, and TITLE-ABS-KEY searches in Scopus, thereby maximizing the precision and relevance of retrieved articles.

Following the search and filtering process, study selection was conducted based on predefined inclusion and exclusion criteria. The reporting process adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR) checklist.27 The PRISMA flow diagram (Figure 1) illustrates the procedures undertaken during screening, the number of studies identified and excluded, as well as the criteria applied throughout the selection process.

All identified studies were imported into the Rayyan AI application to facilitate screening. Titles and abstracts were independently reviewed to assess their relevance to the research topic. During this stage, studies unrelated to health care or educational applications of AI in the context of breast cancer were excluded.

Articles were included in the review if they met the following criteria:

Articles were excluded if they were duplicate publications, not relevant to AI-based patient education in breast cancer, not available in full text or behind a paywall without access, or were editorials, letters, or conference abstracts without sufficient data. Disagreements between reviewers were resolved through discussion until consensus was reached. Although formal interrater reliability statistics (e.g., Cohen κ) were not calculated, all stages involved an independent dual review followed by consensus discussion. The selection process and eligibility assessment followed the PRISMA-ScR guidelines (Figure 1).

The data extraction and analysis stages in this scoping review were conducted systematically to identify patterns, evaluate research trends, and present relevant evidence on the role of AI in breast cancer patient education. After completing the study selection process, data from each eligible study were collected using a structured extraction form to ensure consistency and comprehensiveness.

The extracted data included:

All extracted data were compiled into a synthesis table (Table 3) and analyzed thematically.

The analysis identified recurring themes including:

This structured approach enabled meaningful cross-study comparisons and facilitated the identification of research gaps and opportunities to optimize AI use in breast cancer patient education.

The thematic summary and key findings represent the final stage of this scoping review, aimed at identifying patterns, emerging trends, and existing research gaps based on the analysis of the selected studies. In this phase, the synthesized data were categorized into major themes that reflect the diverse roles of AI in the education of patients with breast cancer. These themes provide a comprehensive overview of how AI technologies have been applied to enhance patient understanding, support decision-making, and improve the overall quality of educational interventions. The thematic analysis also highlights areas where further research is needed to optimize the integration of AI into patient education strategies.

Fourteen studies included in our updated scoping review were conducted across diverse geographical regions, from Europe (Germany, Italy, France, the United Kingdom), Asia (China, India, Taiwan, Malaysia, South Korea), Africa (Egypt), and North America (United States). The United States, China, and Egypt each contributed 2 studies, while Germany, Italy, France, the United Kingdom, Malaysia, South Korea, India, and Taiwan each contributed 1 study.

The research settings were equally varied, spanning academic institutions, hospitals, cancer referral centers, professional conferences, and multi-institutional collaborations. Some studies took place in highly controlled environments, such as RCTs, while others were conducted in real-world settings, such as breast cancer screening centers, online patient communities, and multicenter health networks.

This breadth of geographical and contextual representation underscores the global interest in exploring the role of AI in breast cancer patient education, encompassing both high-income and middle-income countries.

The synthesized studies demonstrated methodological variation in evaluating the role of AI in breast cancer patient education. The research designs included:

This methodological diversity reflects the complexity of evaluating AI, with outcomes ranging from reductions in perioperative anxiety and improvements in patient knowledge to perceptions and acceptability of technology

The populations in the fourteen studies analyzed were highly diverse, encompassing breast cancer patients at various stages of the disease, healthy women undergoing screening, and expert panels evaluating AI outputs. Some studies recruited large samples, such as 718 preoperative patients and 800 women undergoing mammography screening, while others involved only small groups, including four oncology experts or 25 simulated questions.29,31,37 Breast cancer patients constituted the dominant population, whether in the context of chemotherapy, postoperative radiotherapy, or metastatic disease.25,32,33 Other studies included 98 patients evaluating AI avatars and 122 patients engaged in chatbot-based education.34,36 A meta-analysis combined data from 1342 women, providing a broader overview of AI interventions.24

This diversity demonstrates that the benefits of AI can be experienced across different population segments. However, the findings also reveal variations in response: patients with higher educational levels or younger age tended to be more receptive to technology, while those with limited digital literacy remain underrepresented, raising the risk of inequitable adoption.13,25

Synthesis of research findings based on the Table

Based on the synthesis table of research findings, several studies have evaluated the use of digital technology in improving the care of breast cancer patients. Each study demonstrates the role of technology in supporting patient education and enhancing awareness regarding breast cancer management.

ChatGPT and PINK provided more accurate information on the side effects of breast cancer therapy compared to Google, although accuracy varied depending on the type of therapy (chemotherapy, targeted therapy, endocrine therapy). This indicates that AI technology can deliver more precise information, yet maintaining consistency across different platforms remains a challenge.28

AI-based education using QR codes reduced anxiety among patients with breast nodules and decreased their informational needs. This highlights the effectiveness of technology in alleviating preoperative anxiety, indicating the potential of digital solutions to enhance patient experience throughout treatment.30

Most women supported the use of AI in breast cancer screening, although they preferred reports to be generated by radiologists. This suggests a gap between the acceptance of technology and trust in human roles within medical processes.13

The ChemoFreeBot chatbot effectively reduced chemotherapy side effects and improved self-care behaviors among breast cancer patients. This demonstrates that AI-based technologies can support patient self-care, although such interventions require professional oversight to maximize their benefits.33

Other studies further confirmed the positive contribution of AI, including significant improvements in patient empowerment, enhanced comprehension through simplified medical texts, and high satisfaction levels in a systematic review and meta-analysis on chatbot interventions.24,31,34

In contrast, although AI tools such as ChatGPT and PINK may provide more accurate information, misinformation can still occur, particularly in the context of cancer therapy management.28

AI-based interventions demonstrated the effectiveness of technology-driven education in reducing anxiety and improving self-care practices.30,33 However, other studies revealed that ChatGPT’s responses were not always fully accurate, necessitating professional supervision.29,34

Patient diversity plays a critical role in interpreting these findings. For example, higher educational levels were associated with more positive views toward AI use in screening, whereas patients with lower educational backgrounds tended to have a more skeptical perception.13

The digital divide remains a major barrier to the implementation of digital health technologies. While technology can enhance patient education, limited access to digital devices among underserved populations hinders equitable adoption. This underscores the need for digital literacy programs and technology training for socioeconomically disadvantaged women.28,30

Ethical considerations in the use of health technologies, particularly AI and chatbots, must be addressed, especially regarding data privacy and the risk of misinformation. For instance, AI-based interventions can help reduce anxiety and provide medical information, but professional oversight is essential to ensure accuracy and ethical delivery of patient information.13,33

Attention to ethnic and socioeconomic diversity is crucial to ensuring fairness in technology use. For example, while AI avatars have been shown to improve patient communication, they should be culturally tailored to better align with patients’ backgrounds and needs.36 Similarly, evaluations of AI-driven tools such as the Artificial Intelligence Patient Librarian (AIPL) were perceived as usable and reliable, although their evaluation in predominantly affluent and highly educated patients limits generalizability, underscoring the need for equitable design and accessibility across diverse populations.32

The evidence indicates that AI-based education for breast cancer remains in its early stages, with most studies focusing on efficacy and perceptions rather than large-scale adoption.25,33,34 While chatbots and LLMs can improve knowledge, readability, and self-care, variability in accuracy and completeness underscores the need for clinical oversight.29,37,38 Patient diversity strongly shapes outcomes: higher education and younger age are linked to greater benefits, whereas studies in predominantly affluent settings limit generalizability.13,25,32 Cultural tailoring and equitable implementation through digital literacy support, multilingual content, low-bandwidth delivery, and human-in-the-loop models are essential to ensure fairness and trust across underserved populations.30,36

Specifically, AI has shown significant potential in supporting breast cancer patient education. ChatGPT was reported to outperform other platforms in terms of accuracy when providing information about treatment side effects.28 Nevertheless, expert reviews indicated that not all answers had sufficient clinical depth, thereby necessitating validation by medical professionals.29,37 ChatGPT has also been shown to simplify medical texts and improve readability, which is particularly useful for patients with lower educational backgrounds, though oversimplification may risk omitting essential clinical detail.31

One of the main challenges in patient education is simplifying medical information to make it more comprehensible, especially for patients with lower educational backgrounds. A study using a set of 25 simulated breast cancer questions demonstrated that ChatGPT was able to reduce the reading difficulty level of medical texts and enhance comprehension, aligning with international recommendations, including those of the American Medical Association, that health education materials should be written at a lower reading level.31

However, while simplification may improve accessibility, oversimplification risks omitting clinically important details that are necessary for informed decision-making.13 This concern is consistent with expert evaluations, which found ChatGPT responses to be understandable and actionable but sometimes incomplete or slightly inaccurate. Therefore, although ChatGPT shows promise in simplifying language for patients, strict medical supervision is required to balance accessibility with accuracy and ensure that essential clinical information is not lost in the process.37

The integration of AI into breast cancer patient education introduces important ethical considerations. AI avatars in educational videos improved communication, but some patients perceived them as impersonal, raising concerns about the extent to which technology can substitute for human interaction in sensitive health care contexts.36 Additionally, privacy and data security are concerns that may arise when patients interact with AI. While ChatGPT does not directly collect personal data, it may generate or handle sensitive information that could compromise safety if inadequately managed.38 These findings underscore the importance of robust privacy safeguards and continuous professional oversight to ensure that AI-based educational interventions remain both effective and ethically responsible.

Several limitations of the current body of research should be acknowledged. The wide variation in sample sizes—from small expert panels to large-scale patient populations—makes generalization difficult. Study designs were also heterogeneous, ranging from randomized controlled trials and retrospective analyses to surveys and systematic reviews, thus limiting the strength of the evidence presented. Most studies examined short-term outcomes only, leaving the long-term impact of AI interventions on patient quality of life unclear. In addition, the accuracy of AI remains imperfect, with risks of incomplete or misleading information. Finally, patients with limited digital literacy or from marginalized groups remain underrepresented, raising concerns about a widening digital divide.

Future research should prioritize large, multicenter randomized controlled trials and longitudinal designs to evaluate the long-term effects of AI-based interventions on clinical outcomes and quality of life. Standardized outcome measures are also necessary to enable comparability across studies. AI integration into clinical practice must always be supervised by health care professionals to minimize misinformation. Efforts to promote digital literacy are critical, especially among socioeconomically disadvantaged patients, to ensure equitable access to technological innovations. Policymakers should establish clear regulations addressing ethics, privacy, and data security, while developers should design culturally sensitive and personalized educational content to enhance patient engagement and trust.