New York State has renewed its deal to put an AI companion robot in the homes of older adults, citing a 94 percent drop in loneliness. The number comes from a survey the robot’s maker helped design, answered on the robot’s own screen.
In October 2026 the New York State Office for the Aging announced that it would extend its partnership with Intuition Robotics, the company behind ElliQ, a tabletop robot with a lamp-shaped head that talks to the person it lives with. The program began in June 2022. By May 2025, 834 older New Yorkers had received one through their county aging agency, and more than 3,500 had applied. The average user is 75. Eight in ten are women.
The announcement led with outcomes: 94 percent of users report feeling less lonely, 97 percent report better overall wellbeing, 79 percent feel more connected to the world around them, and users interact with the robot dozens of times a day. The state’s director called ElliQ “an essential tool to help combat the loneliness epidemic.” The company’s chief executive described a companion that remembers the older adult, checks in each morning, and encourages walks or calls to family.
A four-year, state-funded deployment in hundreds of homes is the closest thing the American field has to a long-term trial. Which makes it worth reading the way a trial would be read.
Where the 94 percent comes from
The state’s 2026 project report describes its method in two lines: anonymized usage data from the device, plus client responses to the Cobot scale, “the first patient-reported outcome designed to assess companion robots across multiple wellness dimensions.”
The Cobot scale was introduced in a 2024 paper in JAR Life. Its authors include Elizabeth Broadbent of the University of Auckland and Murali Doraiswamy of Duke, two serious researchers in the field, alongside Dor Skuler, the chief executive of Intuition Robotics. Broadbent has been open about the design: the study was “primarily run by the company itself,” with fifteen government agencies, and she describes the work as observational. The next step, she said, would be a randomized controlled trial comparing ElliQ with something else. The state’s 2026 report does not cite one, and reports no comparison group of its own.
So the number is a self-report, on a scale co-written by the vendor, collected through the vendor’s product, from people who applied to receive it, with no control group, and the report does not state how many of the 834 answered. None of that makes the 94 percent false. People who have a robot that greets them every morning may well feel less alone, and the finding that most say so is plausible. It makes the number a satisfaction score, which is a different thing from an outcome.
Smaller details point the same way. The 2026 press release says users average 48 interactions a day, up from 29. The project report the release is drawing on says 41. The 2023 announcement said loneliness fell 95 percent; the 2026 one says 94. The figures move, but they move inside a range that no one outside the partnership has checked.
What the robot actually spends its time on
The most useful number in the report is one it does not highlight. Broken down by activity, the average ElliQ day is 40 percent companionship, 19 percent cognitive games, 15 percent wellness coaching, 11 percent health measurements, 9 percent reminders, 4 percent news and entertainment, and 2 percent social connectedness.
Two percent. The category the state is funding this program to address, connection to other people, is the smallest slice of what the device does. The robot can place video calls, share photos, and send messages; the 2025 caregiver app lets family set goals and receive alerts. Those are the features that reach past the living room. They account for one interaction in fifty.
Set that against the 79 percent who “feel more connected to the world around them” and the gap is the story. Feeling connected to the world and being connected to a person are measured here as if they were the same, and the activity log shows they are not. A device that talks to someone forty times a day will make them feel accompanied. Whether it makes them call their daughter is a separate question, and the one number that could answer it is the one that does not get a headline.
A different second listener
New York is not the only government buying companionship for people who live alone. In Korea, the companion doll reports to a social worker: the state’s interest is detection, and the device’s memory feeds a human service that can send someone to the door. New York’s model is gentler and more private. The robot’s weekly report can be shared with a clinician if the user chooses, the caregiver app is opt-in, and the user controls what family sees.
The two programs have inverted priorities. Korea instruments the companion to make sure a person is reached. New York instruments the companion to make sure the person is engaged. Both call the result connection. Only one of them is measuring the people on the other end.
The question the state did not ask
There is an obvious way to find out whether a companion makes an older person’s life larger, and it is not to ask the companion. It is to count what happens outside it: calls made, visits received, the senior center attended, the neighbor spoken to, over a year, against a group that was offered something else. The research partners already named this as the next step in 2024. A state with 59 aging agencies, 3,500 applicants a year, and a four-year-old program has everything needed to run it.
A measurement that counts the calls and the visits rather than the mood is harder and slower. It is also the only kind that can tell relief from replacement, and the only kind that can fail. Measuring whether an AI makes someone more social means being willing to publish the number when it goes the wrong way. A vendor’s satisfaction score cannot go the wrong way, which is exactly why it keeps getting renewed.
New York has done something most governments have not: it has run a companion program in more than 800 homes for four years. It now owns the largest natural experiment in the field, and 3,500 people on a waiting list who could form the comparison group. The money has been committed again. The question that decides whether it was worth it is still unasked: after four years, do these 834 people have more people in their lives, or one robot?
Sources: New York State Office for the Aging, “NYSOFA Extends Partnership to Provide ElliQ AI Companion Program for Older Adults, Announces New Program Data” (press release, October 8, 2026; quotes from Greg Olsen and Dor Skuler). New York State Office for the Aging and Intuition Robotics, “Transforming Care for Older Adults Across the State of New York,” ElliQ project update (February 2026; June 2024 to May 2025 data: 834 participants, 3,500+ applicants, average age 75, Cobot scale, activity mix, 41 interactions per day). New York State Office for the Aging, press release on the 95 percent figure (August 1, 2023). Broadbent, Loveys, Ilan, Chen, Chilukuri, Boardman, Doraiswamy and Skuler, “ElliQ, an AI-Driven Social Robot to Alleviate Loneliness: Progress and Lessons Learned,” JAR Life (2024). The Robot Report (July 1, 2024; Broadbent on the study design and the need for a randomized trial). Robotics and Automation News (October 10, 2026, program extension).
