Mapping artificial intelligence in older adult care: A bibliometric analysis

Technological evolution of AI in older adult care

The development of AI in older adult care has evolved through several interconnected stages, reflecting a broader transformation from technology-centred solutions toward integrated socio-technical care ecosystems. Rather than representing isolated technological advances, these developments illustrate how innovations in robotics, machine learning, sensing technologies, digital health infrastructures, and generative AI have progressively reshaped care delivery for ageing populations. The evolution of AI in older adult care can be understood through four overlapping phases: assistive technologies and social robotics; intelligent monitoring and predictive care; integrated digital care ecosystems; and generative AI-driven personalised support.

The earliest phase focused primarily on assistive technologies and socially assistive robots designed to support specific care-related tasks. Research during this period emphasised companionship, social engagement, emotional support, and assistance with daily living for older adults living independently or in institutional care settings. Social robots were developed to reduce loneliness, facilitate communication, encourage activity, and enhance psychosocial well-being among older adults experiencing isolation or functional limitations.^17,18 These early applications established an important foundation by demonstrating that successful technological interventions in older adult care depend not only on technical functionality but also on user acceptance, emotional engagement, trust, and meaningful human-technology interaction.

The second phase was characterised by the emergence of intelligent monitoring systems powered by machine learning, wearable sensors, computer vision, and smart-home technologies. This transition marked a shift from interaction-oriented support toward data-driven and predictive care models. Rather than responding only to immediate user needs, AI systems increasingly became capable of anticipating risks through continuous monitoring and predictive analytics. Advances in human activity recognition, fall detection, gait analysis, and health monitoring enabled unobtrusive assessment of physical and cognitive conditions among older adults.^19,20 Deep learning algorithms further improved the accuracy of activity recognition, fall detection, and health prediction models, supporting real-time intervention and preventive care in domestic and community environments.^4,21

The third phase emerged from the convergence of AI, IoT, cloud computing, smart-home platforms, and digital health infrastructure. During this stage, AI systems evolved from standalone devices to interconnected care ecosystems that support ageing in place through continuous monitoring, predictive analytics, remote caregiving, and coordinated service delivery.^5,6 Sensors, healthcare information systems, monitoring platforms, and AI algorithms became integrated within ambient assisted living environments that support both older adults and caregivers. This transformation reflects a movement from device-centred interventions to ecosystem-level solutions in which multiple technologies interact to support personalised and scalable care. Reviews consistently indicate that the convergence of AI and digital health technologies has expanded healthcare systems’ capacity to deliver continuous support to ageing populations.^3,4

More recently, a fourth phase has emerged through the integration of generative AI and large language models into older adult care applications. Unlike earlier systems that primarily focused on monitoring and automation, generative AI technologies enable contextual reasoning, personalised communication, adaptive assistance, and conversational engagement. AI-enhanced social robots and conversational agents increasingly demonstrate the ability to provide companionship, cognitive support, health coaching, and emotionally responsive interaction tailored to individual needs. ²² These developments broaden the scope of AI-enabled care beyond physical health monitoring to include psychological well-being, social participation, health communication, and personalised support. Nevertheless, evidence regarding long-term effectiveness, sustained adoption, safety, and large-scale implementation remains limited. Existing reviews continue to emphasise the need for interdisciplinary collaboration, participatory design, and user-centred development to ensure that technological innovation translates into meaningful care outcomes.^3,12

Importantly, each stage of technological advancement has generated corresponding social, ethical, and organisational challenges. As AI systems become increasingly autonomous, data-intensive, and embedded in care environments, concerns about privacy, transparency, accountability, bias, user autonomy, and digital inclusion grow in significance. From a socio-technical perspective, technological innovation and ethical governance should be viewed as co-evolving dimensions of AI-enabled care for older adults. Understanding the field requires attention not only to technical developments but also to the social systems, care relationships, institutional structures, and governance mechanisms within which AI technologies operate.

Ethical governance and socio-technical challenges

As AI technologies have matured, ethical, organisational, and governance concerns have become increasingly central to older adult care research. Socio-Technical Systems Theory suggests that successful implementation depends not solely on technological effectiveness but also on alignment among users, care organisations, professional practices, and institutional governance frameworks. Consequently, ethical considerations have evolved from peripheral concerns into core determinants of adoption, trust, and long-term sustainability.

Early ethical discussions focused mainly on privacy and surveillance issues arising from data collection and monitoring technologies. The deployment of wearable devices, smart-home sensors, ambient monitoring systems, and IoT-enabled platforms has raised concerns about informed consent, data ownership, data security, and the potential erosion of privacy in domestic and institutional care environments.^11,23 Although continuous monitoring offers benefits for health management, fall detection, and risk prevention, it also raises questions about how older adults negotiate the balance between privacy, safety, independence, and care support.

As AI systems became increasingly capable of supporting decision-making and clinical recommendations, ethical concerns expanded to include transparency, explainability, and accountability. Older adults, caregivers, and healthcare professionals must be able to understand, question, and appropriately use AI-generated recommendations, particularly when such recommendations influence diagnosis, treatment, care planning, or risk stratification. ¹¹ Explainable AI has therefore become increasingly important because algorithmic opacity may undermine confidence in technologies that directly shape health-related decisions and care practices.

Issues of autonomy and dignity have also received increasing attention as AI systems assume more active roles in care delivery. Research on socially assistive robots, intelligent monitoring systems, automated care technologies, and dementia-related applications highlights concerns that excessive reliance on automation may reduce meaningful human interaction, diminish personal autonomy, or contribute to depersonalised forms of care.^9,10 These concerns are particularly significant in dementia care, cognitive assessment, and rehabilitation settings, where users may have reduced capacity to evaluate or challenge algorithmic recommendations and care decisions. ²³

More recently, scholarly attention has expanded toward fairness, inclusivity, digital ageism, and cultural responsiveness. Research demonstrates that perceptions of AI are influenced by cultural norms surrounding ageing, caregiving, autonomy, family responsibility, technology adoption, and human-machine interaction.^13,14 Trust, acceptance, and perceived appropriateness may vary substantially across cultural contexts, suggesting that AI systems developed within one sociocultural setting may not transfer effectively to another. Researchers, therefore, increasingly advocate participatory and inclusive design approaches that involve older adults, caregivers, healthcare professionals, and community stakeholders throughout the development process to ensure alignment with diverse user needs, values, and expectations. ²⁴

Viewed through a socio-technical lens, privacy, transparency, autonomy, fairness, cultural responsiveness, and accountability should not be regarded as isolated ethical concerns. Rather, they represent interconnected governance mechanisms that evolve alongside technological maturity. As AI becomes more deeply embedded in healthcare infrastructure and older adult care environments, ethical governance is a prerequisite for sustainable implementation, public trust, responsible innovation, and long-term societal acceptance.

Bibliometric insights into AI and older adult care research

The rapid growth of AI-enabled older adult care has generated growing interest in bibliometric research to understand the intellectual structure, thematic development, and future directions of the field. Existing studies consistently report substantial growth in publications since the mid-2010s, reflecting the increasing integration of AI technologies into healthcare, ageing research, long-term care, and digital health systems.^5,15 However, recent bibliometric investigations differ considerably in scope, analytical focus, and methodological approach.

Chien et al., ⁵ for example, examined the application of AI, IoT, and edge intelligence in long-term care for older adults using bibliometric, content, and Google Trends analyses. Their findings highlighted the growing prominence of intelligent monitoring systems, digital health infrastructures, and long-term care technologies. However, the study focused primarily on technology adoption patterns and thematic development within long-term care settings rather than on the broader AI-enabled older-adult care ecosystem.

Similarly, Barbosa et al. ²⁵ conducted a bibliometric analysis of service robots within transformative service research. Their work identified major research streams related to service robotics, value co-creation, and service innovation. While this provides valuable insight into socially assistive robotics and service contexts, it offers limited coverage of other AI-enabled applications, including machine learning-based risk prediction, dementia classification, mobility analytics, smart-home monitoring, rehabilitation technologies, and generative AI.

Zhu et al. ²⁶ investigated global trends in AI-enabled healthcare technologies for older adults and identified dominant themes related to smart healthcare, machine learning, and digital health innovation. Although this study provided a broad overview of research development, its emphasis remained largely on publication growth and thematic prevalence rather than on the structural relationships that connect contemporary research fronts and emerging conceptual trajectories.

These studies show that existing bibliometric research has often adopted either technology-specific perspectives or descriptive performance-oriented approaches. As a result, the current understanding remains fragmented across the domains of robotics, digital health, intelligent monitoring, healthcare analytics, cognitive impairment, mobility support, and governance. Moreover, existing studies rarely distinguish clearly between active research fronts and emerging thematic priorities, limiting understanding of how the field is evolving and where future research directions are likely to emerge.

Addressing the current fragmentation in the literature requires complementary bibliometric approaches that capture both current intellectual structures and conceptual priorities. Bibliographic coupling identifies active research fronts through shared reference structures among contemporary publications, whereas co-word analysis reveals dominant and emerging thematic developments through keyword co-occurrence patterns. ²⁷ The integration of these approaches enables a clearer distinction between established domains of scholarly convergence and developing areas of innovation. Such an approach is particularly valuable in AI-enabled care for older adults, where technological capability, user acceptance, clinical utility, ethical governance, and policy development are increasingly interconnected.

Viewed through the lens of Socio-Technical Systems Theory, the evolution of AI-enabled care for older adults reflects a growing interdependence among technological systems, older adults, caregivers, healthcare professionals, organisations, and governance structures. The present study, therefore, integrates bibliographic coupling and co-word analysis to provide a comprehensive and forward-looking mapping of AI research in older adult care. By examining technological, human, clinical, environmental, and governance dimensions together, the study extends existing bibliometric research and offers a richer understanding of how AI-enabled care for older adults is evolving in contemporary ageing societies.

Research design

This study employed a bibliometric research design to map the intellectual structure and conceptual development of AI research in older adult care. Bibliometric analysis enables the systematic, transparent, and reproducible synthesis of large bodies of literature through the quantitative examination of publication metadata, citation relationships, and conceptual linkages among scientific publications. ²⁸ The study followed established bibliometric mapping procedures to ensure methodological rigour in database selection, search strategy development, data cleaning, network construction, and cluster interpretation. ²⁷

Two complementary bibliometric techniques were used to address the study objectives. First, bibliographic coupling was conducted to identify current research fronts by clustering documents that share common references, thereby revealing contemporary thematic convergence and active intellectual communities.^27,29 Second, co-word analysis was performed to identify the conceptual structure and emerging thematic priorities of the field through patterns of keyword co-occurrence.^27,30

The integration of bibliographic coupling and co-word analysis was appropriate because the two techniques provide complementary perspectives on scientific development. Bibliographic coupling is particularly useful for identifying contemporary research alignments because it focuses on shared reference structures among publications. In contrast, co-word analysis reveals conceptual relationships and thematic directions by examining the frequency and co-occurrence of keywords. Together, these approaches enabled a comprehensive examination of both the current intellectual structure and evolving thematic landscape of AI-enabled older adult care research.^27,31

Data source and search strategy

The Web of Science Core Collection (WoSCC) was selected as the sole data source due to its multidisciplinary coverage, rigorous journal selection standards, high-quality citation metadata, and widespread use in bibliometric research.^27,32 Previous methodological studies have shown that WoSCC provides reliable citation relationships, standardised bibliographic information, and robust indexing structures suitable for science mapping and bibliometric network analyses. ³³

The literature search was conducted on 14 June 2026 using the Topic (TS) search field, which indexes article titles, abstracts, author keywords, and Keywords Plus. The following Boolean search string was applied: (“artificial intelligence” OR robotic* OR “machine intelligence” OR “deep learning” OR “intelligent learning” OR “neural network*” OR “expert system*” OR “computer vision” OR “big data” OR “knowledge graph*”) AND (“elder care” OR “elderly care” OR “older adult*” OR “elder service*” OR “elder help” OR “elder support” OR “senior support”).

The search strategy combined AI-related terms with older-adult-care terminology to capture research on AI-enabled care, support, assistance, monitoring, and service provision for older populations. The term older adult was included to reflect contemporary terminology, while terms such as elder care and elderly care were retained because they remain widely used in the existing literature.

The initial search retrieved 8,386 records. The results were refined to include only journal articles, resulting in 6,133 records. Restricting the dataset to English-language publications produced 6,085 records. Because this study focused on literature published between 2004 and December 2025, records published in 2026 were excluded (n = 871). The final dataset comprised 5,214 English-language journal articles published between 2004 and 2025.

The screening and verification process was independently conducted by two researchers. Discrepancies were resolved through discussion and consensus. The inclusion criteria were: (a) English-language journal articles, (b) publications indexed in WoSCC between 2004 and 2025, and (c) publications retrieved using the predefined AI and older-adult-care search terms. Records were excluded if they were not journal articles, were not published in English, or were published outside the specified time frame. The study selection process is presented in Figures 1 and 2.OPEN IN VIEWER

OPEN IN VIEWER

Figure 2.

Annual publication and citation trends in AI-enabled care for older adult research.

Data analysis procedure

All bibliographic records were exported from WoSCC in plain-text format and analysed using VOSviewer version 1.6.20, a widely used software package for constructing and visualising bibliometric networks. ³⁴ Prior to analysis, data cleaning was conducted to standardise author names, merge synonymous keywords, and reduce inconsistencies in terminology. This preprocessing step improved network quality, reduced fragmentation, and enhanced the interpretability of bibliometric maps.^27,35

Performance analysis was conducted to assess the field’s development and scholarly influence. Indicators included annual publication output, total citations, non-self-citations, average citations per publication, articles cited, articles cited excluding self-citations, and h-index. Reporting both citation and non-self-citation metrics provided a balanced assessment of scholarly impact.

Bibliographic coupling analysis was performed at the document level to identify current research fronts in AI-enabled care for older adults. A minimum citation threshold of 135 citations per document was applied, resulting in the retention of 53 publications for analysis. The resulting network generated five bibliographic coupling clusters representing the dominant contemporary research fronts in the field.

Co-word analysis was conducted to examine the conceptual structure and thematic priorities of AI research in older adult care. A minimum keyword occurrence threshold of 104 was applied, resulting in the retention of 60 keywords for analysis. The resulting keyword co-occurrence network generated four thematic clusters representing health vulnerability and frailty; AI-enabled care technologies and human-robot interaction; mobility and functional ageing; and dementia and cognitive impairment prediction.

Following network generation, cluster interpretation was conducted iteratively by examining node composition, keyword relationships, citation patterns, representative publications, and supporting literature. Cluster labels were assigned based on the dominant conceptual themes represented within each network. This approach is consistent with established bibliometric mapping procedures and ensures that cluster interpretations reflect the underlying intellectual structure and thematic development of the research domain. ²⁷

Performance analysis

The final dataset comprised 5,214 publications on AI in older adult care. These publications accumulated 79,858 citations, of which 74,375 were non-self-citations. The dataset achieved an h-index of 102, with an average of 15.32 citations per publication. In addition, the publications were cited by 61,114 articles, or 59,094 after excluding self-citations. These indicators demonstrate that AI-enabled services for older adults have become a substantial and highly influential research domain.

The annual publication trend indicates that research in this field emerged gradually before entering a period of rapid growth from the mid-2010s onward. Publication output increased sharply after 2018, with particularly strong growth from 2020 to 2025. This acceleration corresponds with broader advances in machine learning, deep learning, wearable sensing, smart-home technologies, socially assistive robotics, digital health systems, and IoT-enabled care infrastructures. The citation trend also shows a steep increase during the same period, suggesting that AI-enabled older adult care has gained growing scholarly visibility and influence across healthcare, computer science, robotics, gerontology, digital health, and ethics. Overall, the performance indicators confirm that the field has moved from an emerging research area into a rapidly expanding interdisciplinary domain with strong academic impact.

Bibliographic coupling analysis

Of the 5,214 documents included in the dataset, 53 met the minimum citation threshold of 135 citations per document and were included in the bibliographic coupling network. Bibliographic coupling identifies publications that share common references; therefore, documents positioned close to one another in the network draw on similar intellectual foundations and represent related contemporary research fronts.

Table 1 presents the 10 most-cited publications in the bibliographic coupling network, reflecting both citation impact and structural importance. Citation count reflects the overall scholarly influence of a publication, whereas total link strength indicates the extent to which the publication is connected to other documents through shared references. The results show that Dubno et al. ³⁶ recorded the highest citation count, indicating the strong influence of ageing-related sensory decline and audiometric phenotyping within the broader knowledge base. However, Chen et al. ³⁷ had the highest total link strength, suggesting that deep learning-based clinical imaging and disease severity prediction occupy a structurally central position in the network. Other highly connected publications, including Nunez-Marcos et al., ³⁸ Israel et al., ³⁹ Hausdorff, ⁴⁰ and Nait Aicha et al., ⁴¹ highlight the importance of fall detection, gait dynamics, rehabilitation, and mobility-related risk assessment. In parallel, Čaić et al., ⁴² Johnson et al., ⁴³ and Li and Sikström ⁴⁴ indicate the relevance of service robotics, assistive technologies, and neurocomputational perspectives on cognitive ageing.

OPEN IN VIEWER

Table 1.

Top 10 most cited publications in the bibliographic coupling network.

No.	Documents	Total link strength	Citations
1	Chen P, Gao L, Shi X, Allen K, Yang L. Fully automatic knee osteoarthritis severity grading using deep neural networks with a novel ordinal loss. Computerized Medical Imaging and Graphics. 2019;75:84-92. doi:10.1016/j.compmedimag.2019.06.002.	70	175
2	Nunez-Marcos A, Azkune G, Arganda-Carreras I. Vision-based fall detection with convolutional neural networks. Wireless Communications & Mobile Computing. 2017. doi:10.1155/2017/9474806.	57	286
3	Israel JF, Campbell DD, Kahn JH, Hornby TG. Metabolic costs and muscle activity patterns during robotic- and therapist-assisted treadmill walking in individuals with incomplete spinal cord injury. Physical Therapy. 2006;86(11):1466-1478. doi:10.2522/ptj.20050266.	56	289
4	Hausdorff JM. Gait dynamics, fractals and falls: Finding meaning in the stride-to-stride fluctuations of human walking. Human Movement Science. 2007;26(4):555-589. doi:10.1016/j.humov.2007.05.003.	54	208
5	Čaić M, Odekerken-Schröder G, Mahr D. Service robots: value co-creation and co-destruction in elderly care networks. Journal of Service Management. 2018;29(2):178-205. doi:10.1108/JOSM-07-2017-0179.	46	252
6	Nait Aicha A, Englebienne G, Van Schooten KS, Pijnappels M, Kröse B. Deep learning to predict falls in older adults based on daily-life trunk accelerometry. Sensors. 2018;18(5):1654. doi:10.3390/s18051654.	32	140
7	Johnson MJ, Johnson MA, Sefcik JS, Cacchione PZ, Mucchiani C, Lau T et al. Task and design requirements for an affordable mobile service robot for elder care in an all-inclusive care for elders assisted-living setting. International Journal of Social Robotics. 2020;12(5):989-1008. doi:10.1007/s12369-017-0436-5.	28	181
8	Dubno JR, Eckert MA, Lee FS, Matthews LJ, Schmiedt RA. Classifying human audiometric phenotypes of age-related hearing loss from animal models. JARO: Journal of the Association for Research in Otolaryngology. 2013;14(5):687-701. doi:10.1007/s10162-013-0396-x.	27	860
9	Yang L, Ao Y, Ke J, Lu Y, Liang Y. To walk or not to walk? Examining non-linear effects of streetscape greenery on walking propensity of older adults. Journal of Transport Geography. 2021;94. doi:10.1016/j.jtrangeo.2021.103099.	24	157
10	Li S, Sikström S. Integrative neurocomputational perspectives on cognitive aging, neuromodulation, and representation. Neuroscience and Biobehavioral Reviews. 2002;26(7):795-808. doi:10.1016/S0149-7634(0200066-0).	23	176

Overall, Table 1 suggests that the bibliographic coupling network is shaped by three overlapping intellectual streams: AI-enabled clinical and functional risk prediction, assistive and service robotics for older adult care, and environmental and cognitive determinants of healthy ageing.

Figure 3 presents the bibliographic coupling network generated using VOSviewer. In the network, node size represents citation influence, link thickness indicates the strength of bibliographic coupling between publications, colours represent thematic clusters, and spatial proximity reflects intellectual similarity. Publications located closer together are therefore more strongly connected by their shared reference base.OPEN IN VIEWER

The bibliographic coupling analysis revealed five distinct clusters representing the current intellectual structure of AI research in older adult care. These clusters show that the field is no longer organised around a single technological domain. Instead, it is structured around interconnected research fronts involving predictive health intelligence, assistive robotics, cognitive support, smart environments, rehabilitation, fall prevention, and ethical care technologies.

Cluster 1: Predictive health intelligence and assistive support in older adult care

Cluster 1 represents a research front centred on the use of AI to predict, monitor, and support ageing-related health, cognitive, functional, and environmental outcomes. The publications in this cluster show that AI-enabled older adult care extends beyond robotics and smart-home monitoring to include predictive modelling, medical imaging, dementia progression, sleep assessment, activity recognition, built-environment analysis, and digital inclusion. A dominant theme is the application of deep learning and multimodal analytics to identify health risks, including hip fracture, osteoporotic vertebral fracture, brain ageing patterns, and Alzheimer’s disease progression.^45–49 Other studies examine how built environments and street greenery influence health and walking behaviour among older adults using big data and deep learning approaches.^50,51 The cluster also includes work on assistive robots, domestic robots, and socially assistive robotics, highlighting the importance of acceptance, trust, interaction quality, and user-centred design.^52–55 Importantly, this cluster also raises concerns about measurement validity and digital ageism, emphasising that AI systems must remain inclusive, ethical, and sensitive to older adults’ lived experiences.^56,57

Cluster 2: Assistive robotics, cognitive support, and ageing-in-place technologies

Cluster 2 represents a research front focused on assistive robotics, cognitive support systems, fall-risk prediction, and technologies that support ageing in place. The publications in this cluster show that AI-enabled care for older adults is increasingly concerned with how intelligent systems can assist with daily living, compensate for cognitive or functional decline, and improve independence among older adults. A major theme is the development and acceptance of socially assistive robots and service robots for elder care. Studies examine design requirements, affordability, task suitability, user attitudes, and the practical deployment of robots in assisted-living and care environments.^43,58,59 Related work on conversational digital assistants further highlights the importance of interaction style, reciprocity, and task competence in shaping older adults’ engagement with AI systems. ⁶⁰ Another theme concerns cognitive and functional support. Intelligent reminder systems and cognitive orthotic technologies have been developed to support people with memory impairment, while neurocomputational and neuroimaging studies examine cognitive ageing and the detection of prodromal Alzheimer’s disease.^44,61,62 The cluster also includes wearable sensors and smartphone-based approaches for fall-risk prediction and movement analysis, supporting preventive care and functional monitoring.^63,64 Overall, this cluster reflects the convergence of robotics, cognitive assistance, digital health, and ageing-in-place technologies.

Cluster 3: Fall detection, cognitive ageing, and assistive technologies for functional independence

Cluster 3 represents a research front focused on fall detection, cognitive ageing, assistive technologies, and functional support for older adults. The publications in this cluster show that AI-enabled care for older adults is strongly linked to preventing functional decline and to the development of intelligent systems that support independence, mobility, cognition, and well-being. A major theme is fall-related risk assessment and mobility analysis. Vision-based fall detection using convolutional neural networks and studies on gait dynamics demonstrate how AI and movement science contribute to the early identification of instability, fall risk, and age-related mobility changes.^38,40 Another theme concerns cognitive and sensory ageing. Studies on white matter integrity, functional brain network reorganisation, age-related hearing loss, and AI-supported cognitive assistance show how ageing affects neurological, sensory, and cognitive functioning.^36,65–67 The cluster also includes socially assistive and therapeutic robotics, highlighting the use of robots to motivate physical exercise, support dementia care, and improve older adults’ quality of life.^68,69 Overall, this cluster reflects the integration of fall prevention, cognitive support, sensory ageing, and assistive technologies to promote functional independence among older adults.

Cluster 4: Service robots, smart environments, and Human-AI acceptance in older adult care

Cluster 4 represents a research front that links service robotics, smart-environment intelligence, human-AI acceptance, and environmental determinants of older adults’ well-being. The publications in this cluster suggest that AI-enabled care for older adults is increasingly understood as a service ecosystem in which technological performance, user experience, social cognition, and contextual fit interact. A key theme concerns the role of service robots in care and service networks. Studies show that robots may support value co-creation in elderly care by enhancing assistance, interaction, and service delivery, but they may also contribute to value co-destruction when poorly aligned with user expectations, care routines, or social norms.^42,70 Related studies on AI anthropomorphism and robot service contexts further highlight the importance of perceived humanness, self-congruence, trust, and acceptance in shaping user engagement.^71–73 A second theme involves AI-enabled sensing and environmental intelligence. Smart-home activity recognition, wearable sensors, and machine-learning-based brain-morphology analysis demonstrate how AI supports personalised monitoring, cognitive assessment, and health-related prediction.^74,75 Environmental analytics also show how neighbourhood walkability and visual enclosure can influence older adults’ mental health. ⁵⁰ Overall, this cluster reflects the convergence of service robotics, smart environments, AI acceptance, and context-sensitive care.

Cluster 5: Rehabilitation, fall prevention, and age-friendly mobility environments

Cluster 5 represents a research front focused on AI-enabled rehabilitation, fall prevention, mobility support, and age-friendly environments for older adults. The publications in this cluster show that care for older adults is increasingly concerned with maintaining physical function, supporting safe mobility, and reducing risks associated with frailty, falls, musculoskeletal decline, and environmental barriers. A major theme concerns the use of AI and sensor-based systems for clinical assessment and risk prediction. Deep neural networks have been applied to grade the severity of knee osteoarthritis, while daily-life accelerometry has been used to predict falls among older adults.^37,41 Rehabilitation research also examines robotic and therapist-assisted walking, highlighting the role of robotic technologies in supporting functional recovery and mobility training. ³⁹ A second theme concerns the relationship between mobility, built environments, and healthy ageing. Studies using streetscape greenery, semantic segmentation, and Google Street View images show how neighbourhood walkability and environmental design influence older adults’ walking behaviour and leisure mobility.^76,77 The cluster also includes ethical and user-centred perspectives on robot care, especially for older adults with Alzheimer’s disease and their caregivers.^10,78 Overall, this cluster highlights the convergence of rehabilitation technology, fall-risk analytics, environmental design, and ethical mobility support.

Comparative summary of bibliographic coupling clusters

The five bibliographic coupling clusters demonstrate that AI-enabled older adult care is organised around interconnected research fronts spanning predictive health intelligence, assistive technologies, rehabilitation, mobility support, and human-centred care. Clusters 1, 3, and 5 primarily focus on health monitoring, cognitive ageing, fall prevention, rehabilitation, and functional independence through AI-driven assessment and predictive analytics. In contrast, Clusters 2 and 4 emphasise assistive robotics, cognitive support, ageing-in-place technologies, service robots, smart environments, and human-AI acceptance.

These clusters reveal that the field is shaped by both technological and socio-technical perspectives. The technological dimension centres on machine learning, clinical imaging, sensor analytics, activity recognition, and predictive health monitoring, whereas the socio-technical dimension highlights human-robot interaction, trust, digital inclusion, ethical governance, and age-friendly care environments. Overall, the findings suggest that AI research in older adult care is evolving from isolated technological applications toward integrated care ecosystems that combine predictive intelligence, assistive technologies, rehabilitation, and human-centred implementation to support healthy ageing and quality of life.

Co-word analysis

Co-word analysis was conducted to identify the conceptual structure and emerging thematic directions of AI research in older adult care. Unlike bibliographic coupling, which maps shared intellectual foundations through references, co-word analysis examines the co-occurrence of keywords and therefore reflects the dominant concepts, topics, and research priorities within the field. Using a minimum occurrence threshold of 104, 60 keywords were retained for analysis. The most frequent keyword was “older-adults” (1,186 occurrences), followed by “machine learning” (1,075), “older adults” (889), “dementia” (592), and “health” (538). These high-frequency keywords indicate that the conceptual structure of the field is anchored in older adult populations, machine learning applications, dementia care, and health-related ageing outcomes. Table 2.

OPEN IN VIEWER

Table 2.

Bibliographic coupling clusters on AI and older adult care research.

Cluster	Cluster label	No. of publications	Representative publications	Main focus
1 (Red)	Predictive Health Intelligence and Assistive Support in Older Adult Care	16	Badgeley et al. 45 ; Habes et al. 46 ; Lee et al. 47 ; Toledo et al. 48 ; Tomita et al. 49 ; Wang et al. 50 ; Yang et al. 51 ; Rabbitt et al. 53 ; Smarr et al. 54 ; Wu et al. 55	Predictive health analytics, dementia progression, medical imaging, activity recognition, built-environment health assessment, assistive robotics, digital inclusion, and risk prediction
2 (Green)	Assistive Robotics, Cognitive Support, and Ageing-in-Place Technologies	13	Johnson et al. 43 ; Kachouie et al. 58 ; Pino et al. 59 ; Chattaraman et al. 60 ; Pollack et al. 62 ; Davatzikos et al. 61 ; Li and Sikström 44 ; Howcroft et al. 63 ; Uhlrich et al. 64	Assistive robotics, cognitive support systems, conversational agents, ageing in place, memory assistance, fall-risk prediction, wearable sensing, and functional monitoring
3 (Blue)	Fall Detection, Cognitive Ageing, and Assistive Technologies for Functional Independence	9	Nunez-Marcos et al. 38 ; Hausdorff 40 ; Dubno et al. 36 ; Meier et al. 65 ; Ziegler et al. 67 ; Pollack 66 ; Fasola and Mataric 68 ; Shibata 69	Fall detection, gait dynamics, mobility assessment, cognitive ageing, sensory decline, therapeutic robotics, dementia support, and functional independence
4 (Yellow)	Service Robots, Smart Environments, and Human-AI Acceptance in Older Adult Care	8	Čaić et al. 42 ; Čaić et al. 70 ; Heerink et al. 72 ; Reis et al. 73 ; Alabed et al. 71 ; Bianchi et al. 74 ; Lao et al. 75 ; Wang et al. 50	Service robotics, human-AI acceptance, trust, anthropomorphism, value co-creation, smart-home intelligence, environmental analytics, and context-sensitive care ecosystems
5 (Purple)	Rehabilitation, Fall Prevention, and Age-Friendly Mobility Environments	7	Chen et al. 37 ; Israel et al. 39 ; Nait Aicha et al. 41 ; Yang et al. 76 ; Nagata et al. 77 ; Sharkey and Sharkey 10 ; Wang et al. 78	Rehabilitation technologies, fall prevention, mobility support, osteoarthritis assessment, age-friendly environments, walkability, and ethical robot-assisted care

Table 3 presents the 15 most frequent keywords and their total link strengths. Total link strength indicates the degree of connection between a keyword and other terms in the co-occurrence network. Keywords with high total link strength are not only frequently used but also structurally central to the conceptual organisation of the field. The prominence of terms such as “older-adults,” “machine learning,” “dementia,” “health,” “risk,” “depression,” “Alzheimer’s disease,” and “artificial intelligence” shows that AI research in older adult care is conceptually organised around health vulnerability, algorithmic prediction, cognitive impairment, and technology-supported care.

OPEN IN VIEWER

Table 3.

Top 15 most frequent keywords from the co-word analysis.

No.	Keywords	Total link strength	Occurrences
1	older-adults	3280	1186
2	machine learning	3173	1075
3	older adults	2247	889
4	dementia	2143	592
5	health	1740	538
6	people	1530	442
7	risk	1304	389
8	depression	1066	281
9	aging	1034	390
10	prevalence	984	256
11	mortality	931	285
12	association	913	239
13	alzheimers-disease	910	262
14	artificial intelligence	881	349
15	mild cognitive impairment	876	228

Figure 4 presents the keyword co-occurrence network generated using VOSviewer. In this network, node size represents keyword frequency, link thickness indicates the strength of co-occurrence between keywords, colours indicate thematic clusters, and spatial proximity reflects conceptual relatedness. The clustering algorithm identified four thematic clusters, indicating that AI-enabled older adult care is organised around health vulnerability, intelligent care technologies, mobility and functional ageing, and cognitive impairment.OPEN IN VIEWER

Cluster 1: Health risk, frailty, and population-level ageing outcomes

Cluster 1 focuses on health-related vulnerability among older adults and reflects the growing use of AI to identify, predict, and manage ageing-related risks. Keywords such as health, risk, depression, prevalence, mortality, frailty, disease, disability, sarcopenia, outcomes, and population suggest strong links to epidemiological and clinical research. Recent studies demonstrate the increasing use of machine learning and multidimensional health assessment data to predict frailty, sarcopenia progression, adverse health events, and long-term outcomes among older populations.^79–81 Research has also focused on developing interpretable risk prediction models that support clinical decision-making and early intervention strategies. ⁸² Furthermore, emerging evidence highlights the integration of AI-based frailty assessment into broader geriatric care frameworks to reduce disability, improve resilience, and enhance population-level health outcomes. ⁸³

Cluster 2: AI-enabled care technologies and human-robot interaction

Cluster 2 centres on the technological foundations of older adult care and represents the core AI, robotics, and digital health stream within the field. Keywords including artificial intelligence, technology, deep learning, sensors, systems, social robots, human-robot interaction, elderly care, and fall detection reflect the growing deployment of intelligent technologies to support care delivery. Recent studies show an increasing adoption of AI-powered monitoring systems, smart homes, wearable sensors, and Internet of Things platforms for health surveillance and fall prevention.^84,85 At the same time, social robots and conversational AI systems are increasingly being used to address loneliness, companionship, emotional support, and social engagement among older adults.^86,87 Broader reviews further suggest that AI technologies are becoming embedded within integrated models of geriatric care, extending beyond technical functionality to encompass relational, psychological, and quality-of-life dimensions. ³

Cluster 3: Mobility, physical activity, and functional ageing

Cluster 3 represents a functional ageing and mobility-oriented research stream focused on maintaining independence and physical functioning in later life. Keywords such as physical activity, gait, falls, exercise, balance, walking, mobility, prevention, quality of life, validation, reliability, and validity indicate substantial attention to movement assessment and fall prevention. Recent evidence shows that AI-enabled mobility assessment increasingly relies on wearable sensors, inertial measurement units, and machine learning techniques to detect mobility limitations, assess balance, and predict fall risk.^88,89AI-assisted screening tools have also been developed to improve functional mobility assessment and support individualised intervention planning among older adults. ⁹⁰ Furthermore, wearable technologies are increasingly used to monitor physical activity patterns and promote healthy ageing while providing objective measures of exercise outcomes and quality of life. ⁹¹ Research on fall-risk assessment technologies further demonstrates the importance of valid and reliable digital tools for supporting mobility preservation and independent living. ⁹²

Cluster 4: Machine learning, dementia, and cognitive impairment prediction

Cluster 4 focuses on cognitive ageing and AI-based approaches to dementia diagnosis and prediction. Keywords including machine learning, dementia, Alzheimer’s disease, mild cognitive impairment, diagnosis, prediction, and classification indicate a rapidly expanding research area dedicated to early detection of cognitive decline. Recent studies demonstrate the growing use of machine learning and deep learning models to classify Alzheimer’s disease progression, identify mild cognitive impairment, and improve diagnostic accuracy using neuroimaging, clinical, and multimodal datasets.^93,94 Systematic reviews further show increasing interest in AI-driven prediction models for dementia onset, disease progression, and risk stratification across diverse clinical settings.^95,96 Beyond Alzheimer’s disease, machine learning approaches are also being applied to cognitive impairment associated with other neurological conditions, demonstrating the broad applicability of AI-based diagnostic frameworks. ⁹⁷

Table 4 summarises the four co-word clusters identified in the keyword co-occurrence network.

OPEN IN VIEWER

Table 4.

Co-word clusters on AI research in older adult care.

Cluster	Cluster label	No. of keywords	Representative keywords	Main focus
1 (Red)	Health Risk, Frailty, and Population-Level Ageing Outcomes	19	health; risk; depression; prevalence; mortality; association; frailty; age; disease; adults; risk-factors; impact; symptoms; disability; outcomes; population; life; sarcopenia; management	Health vulnerability, frailty, disease burden, depression, disability, sarcopenia, mortality, and population-level ageing outcomes
2 (Green)	AI-Enabled Care Technologies and Human-Robot Interaction	16	older adults; people; aging; artificial intelligence; care; technology; deep learning; elderly; loneliness; sensors; model; social robots; system; human-robot interaction; elderly care; fall detection	AI-enabled care systems, sensors, deep learning, social robots, human-robot interaction, loneliness, fall detection, and person-centred digital care
3 (Blue)	Mobility, Physical Activity, and Functional Ageing	15	older-adults; performance; physical-activity; gait; falls; validation; reliability; validity; scale; exercise; balance; walking; mobility; prevention; quality-of-life	Mobility assessment, physical activity, gait, balance, fall prevention, exercise, measurement validation, functional independence, and quality of life
4 (Yellow)	Machine Learning, Dementia, and Cognitive Impairment Prediction	10	machine learning; dementia; alzheimers-disease; mild cognitive impairment; impairment; alzheimer’s disease; diagnosis; cognitive impairment; prediction; classification	Dementia diagnosis, Alzheimer’s disease prediction, cognitive impairment classification, machine learning, and early detection

Comparative summary of Co-Word clusters

The four co-word clusters show that AI research in older adult care is organised around interconnected health and technology priorities. Cluster 1 emphasises health vulnerability, including frailty, depression, sarcopenia, disability, mortality, and population-level ageing outcomes. Cluster 2 represents the technological and relational core of the field, focusing on AI, deep learning, sensors, social robots, human-robot interaction, loneliness, and care support. Cluster 3 highlights functional ageing through mobility, gait, falls, exercise, balance, prevention, and quality of life. Cluster 4 focuses on cognitive ageing, particularly dementia, Alzheimer’s disease, and mild cognitive impairment, as well as diagnosis, prediction, and classification.

These clusters suggest that AI-enabled older adult care is moving beyond isolated technological applications toward an integrated care model that combines risk prediction, intelligent care technologies, mobility support, and dementia-related diagnosis. Future research should strengthen links across these areas by developing AI systems that are clinically useful, socially acceptable, ethically governed, and responsive to older adults’ multidimensional care needs.

Socio-technical transition toward integrated care ecosystems

A central finding of this study is the transition from standalone AI technologies toward integrated socio-technical systems of care. Earlier research commonly focused on specific technological applications such as social robots, wearable sensors, fall-detection systems, activity-recognition models, and smart-home devices. In contrast, the current intellectual structure demonstrates an increasing integration of these technologies within broader care ecosystems that involve older adults, caregivers, healthcare professionals, care organisations, digital infrastructures, and regulatory actors. This shift is reflected in both the bibliographic coupling and co-word analyses, in which technological themes are consistently linked to concepts such as frailty, disability, mobility, quality of life, dementia, loneliness, and caregiving.

These findings align with recent evidence showing that AI, IoT technologies, wearable sensing, smart homes, and digital health platforms are increasingly deployed to support ageing in place, remote monitoring, long-term care, and personalised intervention.^3–6 From a socio-technical systems perspective, the effectiveness of AI depends not only on technical accuracy but also on user trust, workflow integration, institutional readiness, and governance structures. Consistent with previous research, acceptance among older adults is influenced by perceived usefulness, ease of use, privacy, trust, autonomy, cultural expectations, and compatibility with existing care relationships.^8,9,14 Consequently, future research should move beyond technological performance metrics and place greater emphasis on implementation processes, caregiver involvement, cultural adaptation, and long-term care outcomes.

Predictive health intelligence, mobility, and preventive care

The findings identify predictive health intelligence and preventive care as major research priorities in AI-enabled care for older adults. Bibliographic coupling revealed strong research streams involving clinical imaging, disease progression modelling, fracture prediction, activity recognition, and built-environment health, while co-word analysis highlighted themes related to frailty, depression, disability, mortality, sarcopenia, and population-level ageing outcomes. These patterns reflect a broader transition from reactive care models toward proactive and preventive approaches. Recent developments in machine learning and deep learning increasingly support the prediction of frailty, functional decline, depressive symptoms, hospitalisation, mortality, and other adverse health outcomes among older adults.^79–83

Closely connected to this trend is the growing focus on mobility, rehabilitation, and functional independence. Research involving gait assessment, fall detection, physical activity monitoring, rehabilitation planning, and age-friendly environments demonstrates how AI is being used to maintain independence and reduce mobility-related risks. Emerging applications utilise wearable sensors, inertial measurement units, computer vision, and machine learning algorithms to assess balance, predict fall risk, and support rehabilitation interventions.^88–92 Furthermore, studies examining walkability, neighbourhood design, and environmental characteristics indicate that AI is increasingly informing age-friendly urban planning and community-level interventions.^50,76,77 Together, these findings suggest that AI should be viewed not only as a clinical monitoring tool but also as a mechanism for promoting healthy ageing, mobility, and quality of life across both healthcare and community settings.

Human-centred AI, dementia care, and adoption challenges

The findings also highlight the growing importance of human-centred AI within older adult care. Multiple bibliographic coupling clusters focused on social robots, conversational agents, human-robot interaction, service robots, trust, anthropomorphism, and user acceptance. Similarly, the co-word analysis identified a major conceptual cluster centred on social robots, loneliness, care, and human-robot interaction. These patterns indicate that the field increasingly recognises that successful adoption depends on how older adults perceive, experience, and interact with AI-enabled technologies rather than on technical performance alone.

This human-centred perspective is particularly evident in dementia and cognitive impairment research. Machine learning, dementia, Alzheimer’s disease, diagnosis, prediction, and classification emerged as a distinct conceptual cluster, reflecting growing efforts to improve early detection and disease management through AI-supported analysis of neuroimaging, behavioural, clinical, and multimodal datasets.^93–97 While these developments offer significant potential for earlier intervention and personalised care planning, they also raise important concerns regarding autonomy, informed consent, privacy, accountability, and human oversight. Ethical analyses consistently emphasise that AI systems used in dementia care must preserve dignity, relational care, transparency, and professional judgement.^11,12,23,98

Cultural factors further complicate implementation. Evidence suggests that attitudes toward AI-assisted assessment vary substantially across countries and populations. For example, Giannouli ⁹⁹ reported that older Greek adults demonstrated limited understanding of AI technologies, expressed substantial distrust of AI-supported assessments, and strongly preferred evaluations conducted by healthcare professionals rather than fully automated systems. Such findings highlight that adoption depends not only on technological capability but also on trust, digital literacy, cultural expectations, and perceived preservation of autonomy. Future research should therefore prioritise participatory design, co-design approaches, long-term deployment studies, and cross-cultural evaluation to ensure that AI technologies remain aligned with the needs and preferences of older adults and their caregivers.

Ethical, legal, and governance implications

A major implication of this study is that ethical governance should be viewed as a foundational component of AI-enabled care for older adults rather than a secondary consideration. Although ethical governance did not emerge as a standalone cluster, concerns about privacy, autonomy, dignity, fairness, accountability, transparency, and trust were embedded throughout the field’s intellectual structure. Studies involving social robots, dementia care, smart-home monitoring, predictive analytics, and AI acceptance consistently emphasise the need for responsible implementation.^{10,11,14,56,98}

The findings also suggest that legal and regulatory considerations are becoming increasingly important, particularly in dementia care and cognitive assessment. As AI systems become more involved in diagnosis, monitoring, decision support, and financial capacity assessment, questions arise regarding consent, liability, explainability, data governance, and professional accountability. Existing dementia care legislation and mental-capacity frameworks in many jurisdictions have yet to fully address the implications of AI-assisted decision-making. Consequently, future research should explore how health data governance, mental-capacity legislation, dementia care policies, and AI regulation can be integrated to protect cognitively vulnerable older adults while supporting innovation.

Overall, the findings indicate that the future of AI-enabled care for older adults will depend on striking an appropriate balance between technological advancement and human-centred implementation. Responsible innovation requires not only accurate and efficient AI systems but also transparency, inclusivity, cultural responsiveness, ethical governance, and sustained human oversight. Future research should therefore prioritise interdisciplinary collaboration among artificial intelligence, gerontology, healthcare, rehabilitation science, public health, law, and ethics to develop integrated care ecosystems that enhance healthy ageing, independence, dignity, and quality of life.

Implications