A randomized waitlist-controlled trial of 274 dementia caregivers in 43 countries found that a fully automated, AI-enhanced self-guided program called PDC30 produced small to moderate improvements in depression (Cohen's d = -0.37), caregiving burden (d = -0.34) and positive aspects of caregiving (d = 0.42) after one month, compared with caregivers who were waiting for access. The trial, led by researchers at the Education University of Hong Kong and Hong Kong Polytechnic University, found no significant effect on anxiety. It was published in the American Journal of Geriatric Psychiatry, posted online January 8, 2026, and appearing in the journal's October 2026 print issue.
The program differs from the clinician-supported app this paper covered in an earlier trial, which eased caregiver distress without measurably reducing symptom severity. PDC30 involves no clinician contact at all: it is available around the clock online and combines a self-help guidebook, an AI-powered counseling chatbot, and interactive tools for cognitive-behavioral techniques, relaxation and activities meant to strengthen the bond between caregiver and the relative receiving care.
Of 441 people who responded to online study recruitment, 274 met the trial's entry requirements, which included providing at least 10 hours of weekly care to a community-dwelling relative with dementia, scoring 5 or higher on the PHQ-9 depression screening questionnaire, passing a basic English comprehension test, and not being enrolled in another caregiver support program. Participants were split by computer-generated randomization into immediate access to PDC30 or a one-month waitlist, with outcomes measured using standard self-report scales for depression, burden, anxiety and positive gains from caregiving.
Because recruitment depended on caregivers finding and responding to an online study notice, passing an English-language screen and navigating a web platform unsupervised, the sample is weighted toward caregivers who already have internet access, English proficiency and some comfort with technology; the authors themselves describe the trial as a test of feasibility and acceptability, not a comparison against in-person or clinician-led care. A separate analysis by the same research team, published weeks later in the same journal, reported that gains in caregiving self-efficacy and positive reappraisal coping, rather than the chatbot interactions themselves, accounted for most of the measured improvement.