FDA Just Redrew the Line for What Counts as a Medical Device — and a Lot of AI Health Products Just Landed on the Free Side

By Andrew

A wearable that measures your blood glucose or blood pressure has, for years, sat squarely in FDA medical device territory — the kind of product that needs premarket review before it can legally ship. As of January 6, 2026, that’s no longer automatically true. Under new FDA guidance, the same category of device can now be marketed as a general wellness product, with no premarket clearance required, as long as it’s positioned around lifestyle and fitness rather than diagnosing or treating disease.

That’s a bigger shift than “FDA relaxes some paperwork.” It moves actual physiological measurement — not just step-counting or sleep-tracking — out of default regulatory territory, based entirely on how a company chooses to market the product.

Two Separate Rule Changes, Same Direction

The January 6 guidance actually updates two distinct FDA policies at once, and both push the same way: less software and fewer devices default into “regulated medical device,” more of them default into “we won’t enforce against this.”

The first is the General Wellness guidance. Non-invasive wearables measuring physiologic parameters — blood pressure, blood glucose, and similar metrics that would previously have triggered device classification — can now qualify as general wellness products if marketed for promoting healthy lifestyle or general fitness rather than for diagnosing, treating, curing, mitigating, or preventing disease. The FDA is exercising enforcement discretion here rather than declaring these products categorically exempt by law, which matters: the exemption rides on marketing language and intended use, not on what the hardware is technically capable of measuring.

The second is Clinical Decision Support software. Under the revised guidance, software that gives a clinician a single, clinically appropriate recommendation falls outside medical device oversight, provided the clinician can independently review the underlying logic, data sources, and guidelines the recommendation is based on. That’s a meaningfully specific bar: not “the clinician theoretically could disagree,” but that the tool’s reasoning has to actually be inspectable, not a black box the clinician either accepts or rejects blind.

Why Generative AI Specifically Matters Here

Both changes existed in some form as regulatory concepts before this guidance. What’s new — and what several legal analyses of the update flagged specifically — is how directly this now applies to generative AI tools. A generative model producing a clinical suggestion is a fundamentally different animal from the rules-based clinical decision support software the CDS framework was originally written around: rules-based logic is inherently traceable step by step, while a generative model’s output is the product of a training process and a statistical inference, not an explicit decision tree. Whether a clinician can “independently review the logic” behind a generative recommendation is a much harder question to answer cleanly than it is for an if-this-then-that rules engine — and the guidance leans toward enforcement discretion rather than resolving that ambiguity strictly.

The practical effect is that a wave of AI-enabled clinical support tools and consumer-facing wellness products that would have needed to seriously consider FDA clearance a year ago now have a plausible, guidance-backed argument for staying outside that process entirely — as long as they’re careful about how the product is positioned and what specific claims it makes.

Where the Line Still Holds

None of this is a blanket exemption. The guidance is explicit that high-risk products — anything that diagnoses or treats disease, or clinical decision support that doesn’t let a clinician independently verify the recommendation’s basis — remains fully subject to FDA device oversight. A blood-glucose wearable marketed to help diabetics manage insulin dosing is a medical device regardless of what this guidance says; the same hardware marketed as “track your wellness trends” is not. The line moved, but it didn’t disappear, and it now runs through marketing and framing decisions as much as through the underlying technology.

What This Means for Anyone Building in This Space

For a company building consumer health hardware or AI-driven clinical tools, this guidance changes the calculus on where the regulatory line actually sits — but it also raises the stakes on getting the positioning right, since the same product can land on either side of that line depending on how it’s marketed and what claims accompany it. A wellness product that drifts into disease-related claims in its marketing copy, or a CDS tool whose “recommendation” isn’t actually reviewable by the clinician using it, can lose the enforcement-discretion protection this guidance offers even without a single line of the underlying software changing. That’s a genuinely different kind of compliance risk than the traditional “did we submit the right paperwork” question — it’s closer to a continuous check on marketing and product framing, not a one-time regulatory clearance.