Voice Assistants: The People Who Need Them Most Use Them Least

Speech needs no screen and no fingertips. So if any group looks built for the talking computer in everyone's pocket, it is the patients who cannot easily read a phone — and they are precisely the ones least likely to have tried it. In a survey of 218 primary-care patients at a single US academic centre, 39.0 percent of those with a visual disability had never used a digital voice assistant, against 24.6 percent of everyone else.

That inversion is the whole point of the paper. Overall appetite was high: 69.7 percent of the sample had used Siri, Alexa or Google Assistant at some point, and 84 percent were open to using one in future, including more than half of those who had never tried. The technology is wanted and largely free. The question the study poses — without claiming to answer it — is why the people it would suit best are furthest from it.

The design, and its limits

This is a needs assessment, not a trial. Over 19 months — recruitment slowed by the COVID-19 pandemic — investigators at the University of Michigan recruited a convenience sample of 218 adults across three clinics: family medicine, a low-vision eye clinic, and physical medicine and rehabilitation. Participants completed a 46-item, investigator-developed questionnaire (shortened to 34 items for the low-vision group) on what they knew about voice assistants, how they used them and what worried them. Of those who answered the relevant items, 36.5 percent reported a visual disability and 30.1 percent a physical one. The instrument was not validated and the sample came from one health system. The authors say all of this plainly; it fixes the ceiling on what the numbers can bear. Read it as a cross-sectional survey published in August 2025 in JMIR mHealth and uHealth, and read the findings as hypotheses.

What the data will bear

Two results sit on firm enough ground to repeat. People with visual disabilities were significantly more likely never to have used a voice assistant (39.0 vs 24.6 percent, P = .03). And among those who had adopted one, they leaned on it far harder: 63.8 percent relied on it sometimes or always, against 32.7 percent of other users (P = .001). The least-served group is also the group that gets the most out of the tool once it is in hand.

The contrast with physical disability sharpens the reading. There, reliance did not differ from everyone else's (45.0 vs 34.9 percent, P = .31), and the higher never-used rate fell short of significance (34.0 vs 21.1 percent, P = .09). The benefit is not generic to disability. It is specific to the visual case, where speech genuinely stands in for a sense the user lacks — which is exactly where you would expect a voice interface to earn its keep, and the most defensible signal in the paper.

Where the evidence stops

A snapshot of a few hundred people at one clinic, answering an unvalidated questionnaire while knowing the study was about voice assistants, cannot say why the gap exists — cost, awareness, earlier frustration, setup hurdles, trust — and cannot show that closing it would improve anyone's health. Worry about privacy, security and confidentiality was common across the whole sample, security most of all, with 54.5 percent moderately to highly concerned (privacy 47.8 percent, confidentiality 51.7 percent). Tellingly, only security concern tracked with using the assistant less often (P = .04); privacy and confidentiality worries did not, and the authors report no general dampening of use by concern. It is a thread worth pulling in a larger study, not a cause established here.

One disclosure belongs in the appraisal, not a footnote: a co-author is the editor-in-chief of the journal that published the study. The authors state it, and it does not sink a descriptive survey. But for evidence this preliminary the standard correction holds — the finding earns its weight when an independent group, sampling more widely, sees the same shape.

The lesson that travels

It is easy to treat availability as access. A tool can be cheap, present on nearly every phone and genuinely useful to a particular group, and that same group can still be the one least likely to use it. For anyone weighing digital health for older or disabled patients in European systems, that is the part of this small study that outlasts its size: do not read uptake off suitability. The people a technology would help most are not automatically the people it reaches, and the gap between the two is where the real work waits.

Source: Rajan M, Furgal A, Kadri R, et al. Exploring the Utility of Digital Voice Assistants for Primary Care Patients, Including Those With Physical and Visual Disabilities: Cross-Sectional Study. JMIR mHealth and uHealth 2025;13:e66185. A single-centre, exploratory cross-sectional survey using a convenience sample and an unvalidated, investigator-developed instrument — descriptive, not generalisable, and with a co-author who edits the publishing journal.