Artificial intelligence (AI) is increasingly applied across the dementia care trajectory to support assessment, monitoring, decision-making and personalized care. Family and informal caregivers are primary end-users and decision-makers for many AI-enabled supports because they deliver most day-to-day care as dementia progresses. Successful real-world implementation of AI-enabled dementia care therefore hinges on caregivers’ willingness to accept, trust and integrate these technologies. Existing qualitative evidence is fragmented across technologies and settings, and prior syntheses have largely examined single modalities rather than the broader landscape of AI-driven tools. This protocol describes a qualitative evidence synthesis designed to integrate caregivers’ lived experiences, perceptions and expectations to inform caregiver-centered development, implementation and governance of AI in dementia care.
The review aims to systematically synthesize qualitative evidence on family and informal caregivers’ acceptance, concerns and expectations regarding AI-enabled dementia care. Specific objectives are to identify factors that influence caregiver acceptance of AI, characterize concerns that may hinder implementation (for example privacy, accountability or reduced human interaction), and describe caregiver expectations for how AI should complement caregiving. The review will also identify evidence gaps to guide future research.
The review will be conducted following the Joanna Briggs Institute (JBI) methodology for qualitative systematic reviews. The protocol was developed and reported in line with the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) 2015 statement. The completed synthesis will adhere to the Enhancing Transparency in Reporting the Synthesis of Qualitative Research (ENTREQ) statement. The protocol is registered with PROSPERO (CRD420261470957). Any important protocol amendments will be documented and reported in the final publication.
Eligibility is defined using SPIDER (Sample, Phenomenon of Interest, Design, Evaluation, Research type):
Sample: family and informal caregivers of people living with dementia (unpaid individuals such as spouses, partners, adult children, relatives, friends or neighbors). Studies focused exclusively on healthcare professionals, paid caregivers or people living with dementia without caregiver perspectives will be excluded.
Phenomenon of Interest: caregivers’ acceptance, concerns and expectations about AI-enabled technologies used in dementia care. Included technologies may encompass machine learning decision support, conversational AI, generative AI, predictive analytics, intelligent monitoring systems and socially assistive robots that incorporate AI. Studies of digital technologies without AI components or conventional telehealth will be excluded.
Design: primary qualitative studies (interviews, focus groups, observations, open-ended qualitative responses). Mixed-methods studies are eligible if qualitative findings are extractable separately.
Evaluation: qualitative findings reporting attitudes, perceptions, trust, concerns, expectations, perceived benefits and challenges related to AI-enabled dementia care.
Research type: peer-reviewed qualitative and mixed-methods studies published in English. Conference abstracts, editorials, commentaries, protocols, opinion pieces, dissertations, reviews and inaccessible full texts will be excluded.
A comprehensive search strategy will be developed with a health sciences librarian, following the JBI three-step approach: an initial limited search to identify keywords and indexing terms, a comprehensive database-specific strategy using controlled vocabulary and free-text terms, and reference list checking of included studies and relevant reviews. The strategy will combine terms for three core concepts: dementia, family and informal caregivers, and AI-enabled dementia care (including conversational AI, generative AI, machine learning, predictive analytics, intelligent monitoring and socially assistive robots).
Electronic databases to be searched from inception to January 2027 are PubMed, Embase, CINAHL, PsycINFO and Web of Science. The focused systematic literature search is planned between 1 October 2026 and 31 December 2026. Only English-language peer-reviewed studies will be included. Backward citation searching will be conducted for included studies and relevant reviews; forward citation searching may be performed using Web of Science. Grey literature will not be systematically searched.
Search results will be imported into EndNote and deduplicated, then transferred to a systematic review management platform for screening. Two reviewers will screen records independently in two stages: title/abstract screening followed by full-text assessment against predefined SPIDER-based eligibility criteria. Reasons for full-text exclusions will be documented. Disagreements will be resolved by discussion and, if necessary, a third reviewer. A pilot screening exercise will assess inter-reviewer agreement and refine criteria prior to formal screening.
Methodological quality of included qualitative studies will be appraised independently by two reviewers using the JBI Critical Appraisal Checklist for Qualitative Research. Any disagreements in appraisal will be resolved through discussion or adjudication by a third reviewer. The appraisal outcomes will inform interpretation of findings but will not be used as an automatic exclusion criterion unless pre-specified in protocol amendments and reported.
Data will be extracted using a predefined extraction form capturing study characteristics, participant details, AI technology types and qualitative findings related to acceptance, concerns and expectations. Thomas and Harden’s thematic synthesis method will be used to integrate study findings: line-by-line coding of findings, development of descriptive themes, and generation of analytical themes that transcend individual studies. If study reports lack clarity, corresponding authors may be contacted for clarification.
Confidence in the synthesized review findings will be assessed using the GRADE-CERQual approach, which evaluates methodological limitations, relevance, coherence and adequacy of data contributing to each review finding. The assessment will support transparent statements about how much confidence stakeholders can place in each synthesized theme.
At the time of protocol publication, the review had not commenced. The planned timeline is: literature searching Oct–Dec 2026; record screening completed by January 2027; data extraction by March 2027; data synthesis and interpretation by May 2027; final review results by July 2027. These dates are indicative and may be adjusted depending on workload and volume of eligible studies. The protocol is registered on PROSPERO (CRD420261470957).
Ethical approval is not required for this review. Findings will be disseminated via peer-reviewed publication. The authors report no specific funding and no competing interests. The review’s outcomes are intended to inform healthcare professionals, technology developers and policymakers on designing and governing caregiver-centered AI solutions in dementia care and to identify gaps for future research on responsible AI integration.