The review, conducted by Keyi Li of The University of Hong Kong, Chengliang Wang of East China Normal University and Australian Catholic University, and Zhuopeng Yu of The Hong Kong Polytechnic University, set out to map a research landscape that has become fragmented across two communities that rarely talk to each other. Computer scientists publish on sensing algorithms and predictive models; gerontologists publish on the lived experience of aging. Because the two literatures barely intersect, it has been genuinely difficult to answer basic questions: which AI technologies exist for supporting older adults at home, how mature they are, and how well they match everyday needs. The team addressed this by conducting a general systematic review with descriptive mapping, searching six major databases, including MEDLINE, CINAHL, Web of Science, ProQuest Central, PubMed, and Scopus, for articles published between 2005 and June 2025.

Following the PRISMA guidelines, the standard protocol for ensuring systematic reviews are transparent and reproducible, the search yielded 101 articles. That number alone tells a story. Research on artificial intelligence for Aging in Place has grown rapidly over the two decades covered by the review, reflecting both the maturation of machine learning and the mounting demographic pressure that gives the field its urgency. But the authors found a persistent disparity at the heart of this literature: computational advancement has consistently outpaced studies on real-world efficacy. In other words, the field is excellent at building systems and comparatively weak at demonstrating that those systems work in the messy, unpredictable environment of an actual older adult’s home.

To bring order to this sprawling literature, the review introduces a four-category taxonomy that classifies AI systems by their algorithmic complexity. At the simpler end sit classical machine learning approaches, which rely on statistical techniques to detect patterns in sensor data, activity logs, or health records. These methods are interpretable and computationally cheap, making them attractive for resource-constrained home environments. Further along the spectrum lie deep learning techniques, including convolutional neural networks, which excel at extracting meaning from raw signals such as camera footage, audio, or wearable sensor streams. Convolutional neural networks, in particular, have become a workhorse for recognizing falls, monitoring gait, and identifying changes in daily routines that might signal declining health.

The taxonomy’s upper tiers capture the newest and most powerful technologies. Large language models, the class of systems behind modern conversational AI, represent a qualitative leap in what home-based support might look like, offering natural-language interaction, personalized reminders, and the potential to serve as always-available companions or assistants. The review also highlights explainable AI, an increasingly important requirement in care contexts where a system’s recommendations must be understood and trusted by older adults, family caregivers, and clinicians alike. When an algorithm flags a risk or suggests an intervention, opacity is not just an inconvenience; it can be a barrier to adoption and a source of harm. Ordering systems along this complexity spectrum allows researchers and practitioners to see, at a glance, where the field’s energy is concentrated and where mature, simpler solutions may already be sufficient.

Beyond classifying the technologies themselves, the authors used topic modeling, a computational technique that identifies recurring themes across large bodies of text, to distill the 101 articles into five application scenarios. These scenarios map the practical domains where AI is being deployed to support Aging in Place, spanning the core needs the review identifies: safety, health, daily activities, and social connection. The scenario structure matters because it shifts the framing from what algorithms can do to what older adults actually require. A fall-detection model, however sophisticated, is only valuable insofar as it addresses a genuine risk an older person faces while living alone. By anchoring the analysis in application scenarios, the review creates a common vocabulary that both engineers and gerontologists can use.

The synthesis culminates in a three-layer organizing framework designed to connect technical capability to gerontological need. This kind of architecture is the review’s most consequential contribution, because it treats AI for Aging in Place not as a collection of isolated gadgets but as a system with distinct levels: the underlying technologies, the application scenarios they enable, and the human outcomes they are meant to serve. The framework is paired with a future research agenda that explicitly targets the disparity the authors documented. The implication is pointed: the next wave of research should not simply chase higher accuracy benchmarks but should demonstrate efficacy in real homes, with real older adults, over meaningful periods of time.

Why does this matter now? The demographic arithmetic is unforgiving. As the review’s authors note, the needs created by Aging in Place already outstrip available care resources, and the imbalance will only intensify as populations age. Artificial intelligence is attractive precisely because of its scalability: software can monitor, detect risks, and personalize support at a marginal cost approaching zero, extending the reach of a care workforce that cannot expand fast enough. A single well-designed monitoring system can watch over a home around the clock; a human caregiver cannot. The economics are compelling, but the review’s evidence base suggests the field must earn that promise through rigorous validation rather than assume it.

There are also quieter lessons in the review’s methodology. By searching six databases across two decades and applying PRISMA standards, the authors assembled one of the most comprehensive maps of this domain to date, and their descriptive-mapping approach, which uses topic modeling alongside traditional synthesis, shows how computational methods can be turned on the literature of computing itself. The work received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors, and the authors declare no competing interests. The article is open access, published on 1 October 2026, meaning its taxonomy and framework are freely available to the researchers, designers, and policymakers who will need them most.

The broader takeaway is a familiar one in technology and aging, but stated here with unusual clarity: the bottleneck is no longer imagination or even raw algorithmic power. It is alignment. Systems must be matched to the rhythms of daily life, evaluated in the environments where they will actually operate, and designed so that the people they serve can understand and trust them. The review’s four-category taxonomy, five application scenarios, and three-layer framework give the field a shared structure for that work. If the coming decade of research follows the agenda these authors lay out, the gap between what AI can compute and what older adults need may finally begin to close, and the goal of growing old in one’s own home could move from aspiration to standard practice.

Subject of Research: A systematic review of artificial intelligence technologies for supporting aging in place

Article Title: Artificial intelligence technologies for aging in place: a systematic review

Article References: Li, K., Wang, C., & Yu, Z. (2026). Artificial intelligence technologies for aging in place: a systematic review. Artificial Intelligence Review. https://doi.org/10.1007/s10462-026-11698-0

Image Credits: AI Generated

DOI: 10.1007/s10462-026-11698-0

Keywords: artificial intelligence, aging in place, gerontechnology, systematic review, machine learning, deep learning, large language models, convolutional neural networks, explainable AI, gerontology, older adults, home care