Does robot pet therapy actually work? The evidence
Robot pets have real, replicated effects on loneliness and positive mood, mostly in care settings and mostly measured over four to twelve weeks, and no demonstrated effect on depression once you restrict the evidence to randomized trials.
Robot pet therapy does something measurable for older adults, and it is narrower than the box implies. Loneliness moves. Positive mood moves. Depression does not, at least not in the two meta-analyses that admitted randomized trials only, where live animals cleared significance and robot pets did not. No trial runs longer than a few months, and the AI-pet tier above $400 has almost no trial evidence at all.
What has actually been measured
| Outcome | Best available evidence | Result |
|---|---|---|
| Loneliness | Meta-analysis, The Gerontologist, 2025: 19 studies, 1,083 older adults | Pooled -0.590, 95% CI -0.919 to -0.261 |
| Positive affect at home | Pilot RCT, Am J Geriatr Psychiatry, March 2026: 50 caregiver and older-adult pairs, 4 weeks | +7.28 points, 95% CI 3.1 to 11.4, p = 0.001 |
| Negative affect | Same trial | No difference between arms |
| Depression | Two syntheses of randomized trials only: Complementary Therapies in Medicine, 2024 and Healthcare, December 2025 | Robot pets missed significance in both (SMD -1.21, 95% CI -2.79 to 0.38); live animal therapy cleared it in both (g -0.72; SMD -2.04) |
| Medication use in dementia units | Randomized block trial, J Alzheimers Dis, 2017: 61 residents, 12 weeks, PARO | Reported reductions in psychoactive and pain medication use, and in pulse rate |
The number everyone quotes, and its fine print
The pooled -0.59 comes from Mehrabi and Ghezelbash in The Gerontologist, 42 effect sizes from literature searched through October 2024. Three details from the full text matter for anyone shopping.
Pet robots were the reference category in the moderator analysis, and neither humanoid robots nor voice assistants differed significantly from them. Nothing in the data says the shape of the machine is what matters.
Setting mattered. Effects ran larger in institutions than in people's own homes: about -0.393 in nursing homes, -0.468 in dementia care centers, -1.887 in hospitals. An older adult living alone at home, which is who most families are buying for, sits in the weakest cell of the analysis.
Egger's test came back significant (intercept -0.825, p < .01), pointing to publication bias, and the authors say plainly that they cannot tell whether the effects persist long term. Read -0.59 as real and somewhat flattered. The outcome-by-outcome version for the whole companion-robot category sits in the wider evidence review.
Against a real dog, the robot loses
Both syntheses that restricted themselves to randomized trials landed in the same place. In the 2024 review, animal-assisted therapy reduced depressive symptoms (g = -0.72) and pet-robot interventions did not reach significance. The December 2025 network meta-analysis ranked 20 trials and put live animals at SMD -2.04 against passive control, robot pets at -1.21 with an interval crossing zero. Its authors call robot pets "a practical alternative when live animals cannot be implemented," which is a fair summary of the case for them.
The cleanest head-to-head is older and small. A Danish trial published in Psychogeriatrics randomized 100 nursing-home residents, median age 85.5, to six weeks of regular visits from a person accompanied by a dog, by PARO the robot seal, or by a soft toy cat. Across depression, cognition and weight, visit type changed nothing. The one difference was sleep duration in week 3, favoring the dog, and it had vanished by week 6.
The live-animal literature is no fortress either: a 2026 evidence map of 29 randomized trials in dementia concluded that study quality is low and heterogeneity high. The robot is losing to a shaky benchmark.
The most useful new study is the one run in living rooms
Almost every trial above ran in a facility. In March 2026 the American Journal of Geriatric Psychiatry published a single-blind pilot RCT that did not. Researchers at the University of Pittsburgh, working with Allegheny County's Area Agency on Aging, recruited 50 caregiver and older-adult pairs from Pennsylvania's Caregiver Support Program. Half received a Joy for All pet immediately, half at the end of four weeks. Half of the older adults in each arm had a dementia diagnosis.
Caregiver-reported positive affect rose 7.28 points in the intervention arm (95% CI 3.1 to 11.4, p = 0.001), adjusted for caregiver age, caregiving load and dementia severity. Negative affect did not differ. Greater engagement with the pet went with higher positive affect.
The authors' interpretation is the part worth carrying home. Caregivers said the pets became a topic of conversation and a focus of social interaction with visitors, so the team concluded the pets "should be considered an environmental stimulus as much as an interaction partner." That reframes what you are buying. The robotic cat is doing part of its work on the people who come through the room.
It also puts the cheap end of the market on the map. Companion cats are $159.99 on joyforall.com, the Golden Pup $179.99, one time, no app, no subscription. Batteries aside, that is the entire cost of the intervention.
PARO, medication, and a caveat about the paper
The most striking single finding belongs to Petersen and colleagues, 2017: 61 residents, average age 83.4, across five secure dementia units, randomized to PARO or standard activities for 12 weeks. The authors report lower pulse rate and reduced use of psychoactive and pain medications in the PARO group.
Two cautions. The published abstract has the anxiety and depression scales rising in the treatment group while the conclusion says stress and anxiety fell, so the scale results as printed are ambiguous and the medication figures are the usable claim. Nobody has repeated that medication finding in a larger sample in the nine years since. PARO also carries no published price and sells by quote through clinical distributors, which puts it outside a family budget.
Weaker designs point the same way with less force. A scoping review of low-cost robotic pets pooled nine studies and found gains in mood, communication and companionship, alongside real concerns: residents mistaking the pet for a live animal, hygiene, and the ethics of encouraging attachment. An uncontrolled 2022 study in residential care found depression and loneliness both improved (p < 0.001) over six weeks.
What program statistics are, and what they are not
New York's Office for the Aging has distributed 31,500 animatronic pets since 2018 and reports that 75 percent of recipients described a reduction in loneliness, plus "a 75 percent decrease in pain" (announcement dated June 3, 2024). The page attributes the findings to NYSOFA itself and names no independent evaluator, there is no control group, and the pain figure is not tied to a published measure.
Distribution at that scale, replicated in more than 30 states, still tells you something a trial cannot: people keep these pets rather than shelving them. Treat it as evidence of uptake rather than effect.
The expensive tier has no evidence behind it
Moflin at $429 (Casio's US launch price) turns up exactly one entry in PubMed: a qualitative study in three Swiss nursing homes, published in German in July 2026, in which staff described using it to calm residents and prompt conversation. No trial, no effect size. A PubMed search pairing Sony's aibo, above $3,000, with older adults or dementia returns five records, the most recent from 2015. Compare what Moflin is and what aibo is on features by all means, but the research does not distinguish either of them from a $159.99 cat.
Full of energy and standing by for a stroke. (@RyuAibo, 2026-03-05)
Trial dosing has been consistent: 15 minutes three afternoons a week for 10 weeks in the Australian PARO studies, up to 30 minutes three times a week for six weeks in a 2026 Queensland implementation study. Nothing in the published record follows a robot pet for a year, and nothing tests a $160 pet against a $429 one.
If you are deciding for a parent, the practical version is short: start at the cheap tier, put the pet where the person actually sits, and expect part of its value to come from what visitors say about it. The buyer's guide for older adults covers the rest of the category. Before spending anything, call the Eldercare Locator on 1-800-677-1116, which routes to your local Area Agency on Aging. Many of those agencies run the same animatronic-pet programs New York does, and hand the pet over for nothing.
One email when something actually changes: a delivery verified, a price cut, a robot discontinued. No weekly filler, unsubscribe whenever.