A wellness label cannot settle what a health app promises

CAP’s guidance on health apps published on 8 October 2026 draws attention to the boundary between recording information and software that measures, analyses or interprets physiological data. Separately, the ASA’s Do Health ruling on 7 October 2026 examined unsupported impressions about blood testing, disease prevention and personalised advice.

A feature that interprets a blood result changes more than the convenience of an app that previously stored it. The intended purpose may need reassessment, and claims about the resulting advice need their own evidence. Describing the change as better personalisation leaves both questions open.

CAP’s guidance concerns when software may qualify as a medical device. The Do Health ruling concerns what consumers were led to believe about a service. Establishing the product’s regulatory status would not, on its own, substantiate every benefit advertised for it.

Classification starts with the actual purpose

CAP’s guidance discusses three app rulings. In one, there was no clear evidence of medical-device functionality, yet the advertising still raised substantiation issues. A marketer can therefore have claims to resolve even when the question of device qualification has not been established.

The MHRA’s software and apps guidance provides the relevant starting point for examining intended purpose. Its application to Great Britain also needs to be distinguished from the different regime in Northern Ireland. A product’s regulatory position cannot be determined simply by finding the words “AI” or “personalised” in its feature list.

Suppose an app stores blood-pressure readings entered by a user. An update then interprets those readings and recommends action concerning a medical condition. The original wellness positioning would provide little basis for assessing that added function. The team would need to examine its purpose and behaviour before deciding how it could be described.

Marketing is relevant because it helps describe that intended purpose and shapes consumer expectations. A feature may be described internally as an informational aid while advertising presents it as a way to identify a condition. The team needs to resolve the difference between those accounts before the creative is released.

A restrained description cannot resolve the mismatch either. Reviewers need to see the functionality and its intended use, not just the terms the marketing team has chosen for it.

More personal information does not establish a better recommendation

Do Health promoted a health app incorporating a blood-testing programme. The ASA upheld two issues concerning impressions that the service could diagnose or prevent named medical conditions and offer more specifically tailored advice than existing guidance, without sufficient supporting evidence. It also treated a co-founder’s LinkedIn post promoting the venture as within the CAP Code’s scope.

The ruling did not classify Do Health’s software as a medical device or prohibit personalised health advice. It concerned benefits the ASA considered unsupported. Having collected a person’s biomarkers does not itself establish that the advice produced from them is more appropriate or effective for that person.

A hypothetical service might combine a questionnaire with laboratory results and produce advice about sleep, diet and exercise. It could genuinely use personal information. A stronger claim that it knows precisely what the individual needs would require evidence that the recommendations are appropriate to that person. Collecting more information does not, on its own, answer that question.

The Exhale Coffee ruling also concerned evidence that did not support the advertised health proposition. In a testing service, general evidence for healthier behaviour would leave open whether this programme identifies anything useful that the individual should do differently.

Review the story created by the whole communication

A founder’s account of illness can explain why a service exists. Placing that account beside a testing offer may also suggest that the service could help someone avoid the same experience, even without an explicit promise of prevention.

For reviewers, the transitions are important. A reference to a serious condition followed by a testing demonstration can imply a link between the two. A colour-coded result or testimonial may reinforce that inference. The complete communication can make a stronger claim than its individual components, even where each has been reviewed separately.

The founder’s post may also say something different from the advertisement or destination page. Reading the connected customer journey can reveal a promise that is less apparent when those items are reviewed separately.

A founder’s post can promote the service without a purchase button or a formal advertising tone. In the Do Health case, the ASA examined the content and its connection with supplying the service. The account being personal did not settle whether the post was marketing.

The personal story can remain, provided the team examines the health benefits it suggests and checks the evidence for them. Sincerity about why a service was founded does not substantiate what it can achieve.

Product changes need a route back to claims review

A one-off review is vulnerable when the application changes. A new feature might interpret data differently, expand the intended audience or change the advice presented. Marketing may reuse approved copy without recognising that the relationship between product, evidence and intended purpose has moved.

Product teams could flag material changes for claims review, supplying the version assessed and the evidence behind the benefits being advertised. A reviewer could then see whether the earlier approval still covered the function, rather than having to discover the update from a revised screenshot.

An earlier approval is useful only if its conditions still fit the system in use. It may have concerned a narrower function or a different audience. Retaining those details makes reassessment possible, although the quality of the evidence still has to be judged.

For AI-enabled features, that review should avoid treating a change in model as automatically decisive or automatically irrelevant. Its importance depends on what the change does to the product’s function and the claims being made. The assessment needs the product and clinical context as well as technical information.

When a product starts interpreting information it previously only stored, its marketing deserves another look. The business needs to be able to justify the advertised benefit for that version of the service. Calling it wellness will not answer a customer who understood the promise as help with a medical condition.