Summary
Consumer smartwatches, rings and health apps can make studies easier to run, but a Nature interview highlights risks involving data privacy, security and opaque algorithms. These concerns become more complicated when wearable data are combined with AI systems that shape health interventions.
Consumer smartwatches, rings and health apps are making it easier for researchers to collect information about steps, heart rate, sleep and physical activity. But the commercial systems behind those measurements can introduce ethical and scientific risks, according to Wuyou Sui, a digital-health and behavioural-change researcher at Western University in Canada, in a Nature interview published on 16 September 2026.
Sui identifies data privacy and data security as the most pressing concerns. Researchers using a commercial wearable or health app may be working with a third-party developer whose commercial interests and data practices are outside the research team’s control. The resulting information can reveal daily habits and aspects of a person’s physical status, creating risks related to sharing, sale, advertising, insurance and theft.
The interview also describes a separate problem: researchers may not be able to assess how a device or app converts sensor readings into reported measures such as step counts.
Why researchers are turning to consumer devices
Traditional research wearables can be expensive and inconvenient to use. Sui describes earlier work with inclinometers, devices attached to the thigh to identify whether someone is sitting or standing. Participants had to shave part of the thigh and attach both the inclinometer and a waterproof cover.
From 2019, Sui and colleagues began exploring whether step counts from commercial health apps could be used in Pathverse, a digital platform for delivering customised health interventions. In separate work, they collected data from commercial virtual-reality headsets with built-in accelerometers to corroborate participants’ activity levels.
The convenience and popularity of consumer devices can therefore expand the kinds of studies researchers can conduct. Their use also changes the data chain: information may pass through a company’s app, cloud service or algorithm before it reaches the research platform.
Opaque algorithms can affect research conclusions
Many wearable companies do not disclose in detail how their AI-based algorithms process inputs and produce outputs. That makes it difficult for researchers to independently assess the accuracy of a reported measure or understand how it was generated.
Sui gives step counts as an example. If the value shown to a participant is inaccurate, a study platform could use that value to deliver an inappropriate recommendation, such as telling the participant to double their target. The concern is not only whether a device is accurate in a general sense, but whether the measurement is suitable for the particular research question and intervention.
A commercial device maker may also have access to the underlying data under its service terms. Sui says such data could be used to target articles or advertisements, made available to advertisers, or used to train algorithms designed to influence consumer behaviour. Those possibilities can affect privacy and may also complicate the research process by introducing commercial influences around participants’ behaviour.
AI-based interventions add another layer
Wearable data can be combined with artificial-intelligence tools to tailor health interventions. For example, a chatbot could ask a participant about step goals or barriers to exercise, while the intervention system combines those answers with information from the wearable.
Sui contrasts systems developed and controlled by researchers with commercially available tools such as ChatGPT. A research team can build its own large language model with greater oversight, although doing so is harder. Commercial systems may be easier to use, but researchers have less visibility into how they operate and collect data. Detailed prompts can guide their interactions, but prompts do not give a research team full control over the underlying model or its data practices.
The interview presents consumer wearables as useful research tools rather than inherently unsafe devices. Their responsible use requires researchers to account for the accuracy and provenance of measurements, the commercial parties handling the data, and the way AI systems use information to shape participant interactions.
The source is an expert interview, not a new device-validation study or clinical trial. It does not quantify how often wearable-data errors alter research findings or establish which governance controls are most effective. Those questions will depend on the device, app, study design and data arrangements involved.