Scoping reviews assess the extent of the research, range, and nature, identify gaps, determine systematic review value, and disseminate research findings [32]. A scoping review methodology is used to systematically map available research on the broad, complex, and emerging research question [32]. In emerging research fields, a lack of randomized controlled trials may impede formal systematic reviews or meta-analyses. Scoping studies can address diverse questions and incorporate a range of study designs, making them ideal to complement clinical trial findings. The scoping review was conducted following the Arksey and O’Malley framework [32] and was reported in accordance with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines [33] and was structured according to the following outlined steps.

The research question was identified after an initial review of the literature and a discussion within our research team. The key research questions that guided the review were as follows:

Before searching, the research team worked collaboratively in making decisions about inclusion and exclusion criteria and in planning the initial search strategies to comprehensively identify the relevant literature. The final search strategies were developed with the assistance of a health science librarian (Multimedia Appendix 1). The literature search was performed by 3 researchers in October 2024, covering 6 English databases (PubMed, Medline, Embase, Web of Science Core Collection, IEEE Xplore, and CINAHL) and 4 Chinese databases (SinoMed, CNKI, Wanfang, and VIPC). Relevant studies were searched from January 2010 to October 2024 to answer the above research questions (1-6). An example of the search strategy performed in Web of Science Core Collection is presented here: (recommender system* OR hybrid recommendation* OR collaborative filtering OR content based recommendation* OR recommendation* system* OR knowledge based recommendation*) AND (recipe* OR diet* OR food OR eat* OR nutrition*). As the search proceeded, additional terms were suggested by experts to potentially modify the question, but new research did not result in additional data, and the question did not change. Duplicate references were filtered out using EndNote. In addition to database search, hand searching was conducted by screening the reference lists of all included articles and relevant review papers to identify additional eligible studies. We also manually checked key journals in the field and conference proceedings to ensure comprehensive coverage.

All studies searched were independently assessed by 2 authors (XD and BY) based on the inclusion criteria listed below, and discrepancies were verified by a third author (JW). Studies were eligible for inclusion in this review if all the following criteria were met: (1) the recommended information was related to at least one of the following: food, meal plan, diet plan, and recipe; (2) the study applied personalized recommendation strategy; (3) the recommendations were generated using algorithmic and technological methods; (4) the study population comprised patients with chronic health conditions; and (5) the study was published in a peer-reviewed journal or conference proceeding. The exclusion criteria included: (1) the recommendation unrelated to human health (eg, study focusing on animal health [34]), (2) studies reporting the same RSs were considered duplicates; only the most recent or most comprehensive publication was retained (eg, the latest published study [35] was included, while the earlier one [36] was excluded), and (3) full-text articles not in English or Chinese language.

Decisions regarding the information to be recorded from the primary studies were made through discussions within the research group. Subsequently, a structured chart was developed to collate, summarize, and share the extracted data. A descriptive-analytical narrative approach was used to extract and chart the data from the selected articles [32,37]. Three researchers (XD, BY, and ZZ) independently extracted the data and performed coding, with the other two authors (HN and JW) verifying accuracy. All discrepancies were resolved by consensus. The following details were documented for each included study to answer the research questions: (1) nationality and publication year (Multimedia Appendix 2); (2) target users; (3) function structures; (4) recommendation content; (5) recommendation method, recommendation technology, data of training set, and recommendation process; and (6) evaluation method, evaluation criteria, test set, or evaluation sample size.

The scoping review methodology aimed to summarize the breadth and depth of the existing literature. At this stage, an overview of the characteristics of all included articles was collated, summarized, and reported. Initially, a basic numerical summary of the studies, including the extent, nature, and distribution of the articles, was presented. As this was a scoping review, the critical appraisal of the quality of the included studies was not conducted. However, efforts were made to map the diversity and variety of diet-related HRSs based on factors such as the characteristics of target users, function structure, recommendation content, and other factors. This process facilitated researchers in reaching conclusions about the key characteristics of research in this field and provided insights for future studies.

In total, 4492 published studies were identified in the searching process (Figure 1). EndNote X9 was used to exclude 350 duplicates, and 4142 studies were excluded based on a review of their titles and abstracts. The remaining 208 studies were searched for full text. Ultimately, 193 were excluded based on the exclusion criteria, and 15 studies [35,38-51] were included and analyzed.

The publication years and the countries of the included studies are shown in Multimedia Appendix 2. The studies were published between 2010 and 2024, and there has not been a noticeable increase in their publication over the past 5 years. The articles originated from 9 countries (based on the first author’s affiliations), with the top 3 being China (n=5, 33.33%) [39,40,43,44,49], India (n=2, 13.33%) [41,47], and Pakistan (n=2, 13.33%) [38,48].

Among the 15 included studies [35,38-51], only 2 reported the age of the target users, which were 18‐80 years [50] and older than 65 years [39], respectively. The target users and patients’ chronic health conditions are shown in Table 1. Diet-related HRSs primarily target adults with chronic diseases. Among chronic diseases, diabetes and hypertension were the most commonly targeted by diet-related HRSs, with 9 systems (60%) including users with diabetes and 6 (40%) including users with hypertension.

Among the 15 studies included [35,38-51], 9 (60%) [35,38-45] studies described function structures. The function structures described in these studies varied, but certain recurring components were consistently noted, which could be summarized into 4 major components: user information, food or diet recommendations, knowledge and decision support, and data management and additional functions. Detailed information is shown in Table 2.

The recommended content was divided into 5 categories. Detailed information is shown in Table 3.

Table 4 presents detailed information on the implementation of recommendation features in the reviewed diet-related HRSs. Recommendation methods were inductively classified into 3 categories: constraint-based, preference-based, and hybrid. In this review, constraint-based refers to recommendations based on the patient’s condition-related dietary restrictions, while preference-based refers to recommendations based on the patient’s dietary preferences. Hybrid refers to methods that incorporate both constraint-based and preference-based approaches. Among the 15 included studies [35,38-51], 6 (40%) used constraint-based methods [41,43,46,47,49,51], 5 (33.33%) applied preference-based methods [39,40,44,45,50], and 4 (26.67%) adopted hybrid methods [35,38,42,48].

While the recommendation technologies were coded according to terms reported in the studies. The recommendation technology included: KG (n=2, 13.33%) [39,44] and HyR (n=13, 86.67%) [35,38,40-43,45-51].

For the data of the training set, 13 studies (86.67%) [38-42,44-51] explicitly mentioned the data sources of the training set, while the other 2 studies (13.33%) [35,43] did not specify this information. The data sources used in the reviewed studies varied widely and can be categorized into 4 main types: authoritative government and institutional databases (n=5, 33.33%) [38,45,46,48,49], academic databases and publicly available datasets (n=2, 13.33%) [39,41], expert and hospital data (n=3, 20%) [40,47,51], and recipe and user-generated data (n=3, 20%) [42,44,50].

The recommendation process followed a structured workflow, integrating advanced technologies to generate more accurate, adaptive, and context-aware dietary recommendations. The structured workflow included user profiling, integration of structured knowledge (such as food databases or ontologies), personalized filtering and matching based on health conditions and preferences, and ranking of suitable options for final recommendation. While this core process was consistent, 6 systems incorporated advanced technologies to enable more accurate, adaptable, and context-aware dietary recommendations, such as ontology-based reasoning [51], optimization algorithms (eg, ant colony [48]), dynamic modeling using Long Short-Term Memory networks [44], expert rule systems [39,49], and the use of KGs such as FoodKG [39].

Among the 15 included studies [35,38-51], 12 studies (80%) [35,39-44,46-49,51] evaluated the diet-related HRSs, while the remaining 3 studies (20%) [38,45,50] did not conduct an evaluation. Table 5 presents detailed information on the evaluation criteria, evaluation method, and test set or evaluation sample size of the 12 studies [35,39-44,46-49,51].

The evaluation criteria were extracted directly from the included studies, including: accuracy (n=9) [35,41,42,44,46-49,51] refers to the extent to which the system’s predictions of patients’ preferences or nutritional needs match their actual dietary preferences or intake; usability (n=3) [35,42,46] refers to the ease with which patients can, with little or no assistance, successfully operate the system to receive recommendations, navigate, and interact with its interface; acceptance (n=1) [46] refers to the degree to which patients are willing to receive and use the dietary recommendations provided by the system; satisfaction (n=1) [40] refers to patients’ subjective sense of contentment or evaluation of the overall usefulness and experience of the recommendations, measured through participants’ comprehensive comparison of system-generated meal plans with manually designed ones; efficiency (n=1) [40] refers to the amount of time, steps, or cognitive effort required for patients to obtain appropriate dietary recommendations from the system; feasibility (n=1) [43] refers to the practicality of deploying the recommendation system in real-life settings; and effectiveness (n=1) [39] refers to the actual positive impact of the system in real-life contexts on patients’ dietary behavior changes or health outcomes. Four studies [35,40,42,46] included more than one evaluation criterion.

The evaluation methods included: online evaluation (n=7) [35,39,40,42,43,46,49] and offline evaluation (n=5) [41,44,47,48,51]. Among the 7 studies [35,38,42,44,45,48,51] that used online evaluation, 6 studies [35,39,40,42,43,49] evaluated the diet-related HRSs with target users of patients, while 3 studies evaluated the HRSs that relied on experts, such as nutritionists [40,46] and dietitians [49]; their main roles were to validate the recommendation results. Four of the 5 offline evaluation studies [39,40,43,46,47] used the same dataset for training and evaluation [41,44,47,48].

This scoping review revealed that diet-related HRSs for individuals with chronic conditions were still in their early stages, with limited patient-specific designs and significant room for improvement. The review highlighted substantial gaps in target users, system functions, recommendation content, recommendation feature implementation, and evaluation approaches, which need to be addressed to support the development of more effective and patient-centered diet-related HRSs.

Future research should prioritize developing diet-related HRSs based on a user-centered design framework. These systems must be tailored to patients’ specific needs with a deep understanding of their cognitive abilities, digital literacy, cultural dietary preferences, and daily routines. Diet-related HRSs need to address clinical challenges through contextually relevant functional designs, customized for different patient groups, from older adults with multimorbidity to pediatric patients requiring caregiver-mediated support. Practical usability features, such as adaptive reminders, contextualized recipe guidance, and dynamic feedback loops, are essential for real-world effectiveness.

To support patients’ sustained dietary behavior change, recommendations should move beyond static food lists toward evidence-based, actionable guidance using practical tools such as portion guides, ingredient lists, and meal preparation videos [79]. Future systems should adopt hybrid, adaptive recommendation mechanisms and integrate with electronic health records (EHRs) and mobile apps to enhance personalization, interpretability, clinical relevance, and engagement, which is essential for sustaining long-term adherence to dietary changes. Additionally, incorporating behavior change theories, such as Social Cognitive Theory [80], Theory of Planned Behavior [81], and the Fogg Behavior Model [82], can strengthen the dietary intervention design. Gamification elements informed by these theories, such as rewards and rankings, may further motivate users and increase adherence [83].

When managing sensitive health data, especially when integrating with EHRs and personal medical records, the systems must implement robust data governance, including secure storage, encryption, role-based access, and transparent consent. Compliance with regulations such as the US Health Insurance Portability and Accountability Act (HIPAA) [84], the European Union’s General Data Protection Regulation (GDPR) [85], and national standards are essential to safeguard patient privacy and foster user trust. Finally, standardized evaluation methods and robust frameworks, combining clinical trials, behavior change evaluation, and usability evaluation, are essential to verify effectiveness, accuracy, and long-term impact in real-world settings, while ensuring cost-effectiveness and compliance (eg, US Food and Drug Administration approval and European Conformity marking) for safe and scalable healthcare implementation.

For clinicians and dietitians, diet-related HRSs can serve as supportive tools to deliver personalized, evidence-based, and adaptive dietary guidance aligned with patients’ medical conditions, treatment regimens, cultural preferences, and daily routines. By incorporating medication-food interaction checks and actionable educational content, these systems can enhance patient safety, dietary adherence, and long-term health management. Future diet-related HRSs could integrate seamlessly into clinical workflows and demonstrate measurable health outcomes. Embedding diet-related HRSs within EHRs would ensure that dietary recommendations are consistent with medications, lab results, and care plans. By linking dietary adherence to physiological indicators such as hemoglobin A_1c, blood pressure, or lipid levels, systems can enable real-time feedback and timely clinical intervention. Beyond accuracy, future evaluations should predefine clinical end points (eg, hemoglobin A_1c reduction or blood pressure control) and validate effectiveness through randomized controlled trials and real-world studies.

This study has several limitations. First, only studies published from January 2010 to October 2024 were included, which may not reflect recent developments. Second, only English and Chinese literature were reviewed, excluding studies in other languages. Third, the focus was primarily on chronic health conditions, with limited exploration of diet-related HRSs for acute conditions or general wellness, suggesting a need for broader research in future studies. In addition, the included studies themselves demonstrated certain methodological limitations, such as small sample sizes, reliance on offline testing, and limited evaluation of actual behavior change. These issues reflect not only the limitations of this review but also highlight gaps in the existing research that warrant more rigorous clinical evaluation.

Diet-related HRSs can offer personalized dietary recommendations for patients with chronic conditions. The analysis reveals significant gaps, demanding that future systems be grounded in user-centered design to meet patient-specific needs. These systems must recommend more practical and actionable dietary guidance. The adoption of hybrid recommendation techniques can enhance the precision and clinical relevance of dietary recommendations. Establishing standardized evaluation metrics and conducting real-world studies with long-term follow-up will be essential. This will verify the system’s ability to positively change dietary behaviors and improve clinical outcomes. Addressing these issues will transform diet-related HRSs into trustworthy and impactful tools for managing chronic diseases and delivering patient-centered care.