What the Robot Studies Agree On

Every family staring at a companion robot's product page asks the same question, does this thing actually work. There are now enough studies to answer it, if you stack them instead of reading them one at a time. A single study is a rumor, and a pile of single studies is a louder rumor. Two independent meta-analyses, five years apart, are the closest thing this field has to a fact.

Here is what they agree on. A 2021 meta-analysis in Innovation in Aging, pooling randomized trials from the Paro era of robot pets, found agitation meaningfully reduced and depression improved, with nothing clear on quality of life. A 2026 meta-analysis in The Gerontologist, covering the newer AI-based social robots, found depression reduced, agitation reduced by almost the same amount, and nothing significant on cognition or anxiety. Different decades, different hardware, near-identical answer. The robots help mood. The effects are small, and they replicate.

Then comes the loud exception. A 2025 randomized trial of the parrot-shaped PIO robot in day-care centers found the opposite half of the story, cognition scores rose almost four points against a flat control group while depression did not budge. One trial, sixty-six people, six weeks. It might be the start of something or it might be noise, and the honest read is that mood has two meta-analyses behind it and cognition has one small trial. Our own reporting on Korea's Hyodol, the most-deployed care robot in the world, adds the caveat that benefits measured at three months were gone by six.

And there is now a study of who says yes in the first place. Among 506 older Koreans, acceptance came down to confidence and curiosity rather than age, income, or how much help a person needed, and the healthier, more independent elders were the more willing ones. So the full picture reads like this. A companion robot modestly helps the mood of the person who agrees to have one, for as long as the effect holds, and the households that need help most are the least likely to say yes. Buy one as a mood support with a return window, never as a memory treatment, and know that the science, for once, mostly agrees.