What Changed in FDA's 2026 Clinical Decision Support Guidance

On January 6, 2026, the U.S. Food and Drug Administration reissued two of its long-standing digital-health guidance documents, and the short version is this: more clinician-facing software and more connected wearables now sit in a zone the agency does not actively regulate as medical devices. One document covers Clinical Decision Support (CDS) software; the other, titled General Wellness: Policy for Low Risk Devices, covers consumer wellness products. Neither rewrites the law. Both change how the FDA reads the law it already has.

That distinction matters whether you are building, buying, or using these tools, so this explainer walks through what actually moved. It is educational and not medical or regulatory advice.

Key points#

• The underlying statute (Section 520(o)(1)(E) of the Federal Food, Drug, and Cosmetic Act) is unchanged; only the FDA's interpretation of it shifted.
• CDS software that offers a single recommendation may now qualify for enforcement discretion when only one option is clinically appropriate.
• Time-critical use and risk-score outputs are no longer automatic disqualifiers.
• Wearables that estimate heart rate variability, oxygen saturation, glucose, or blood pressure may count as general wellness products under specific conditions.
• Enforcement discretion is not clearance or approval; the FDA can withdraw it.

First, a word on what "enforcement discretion" means#

Much of the confusion around these documents comes from treating three different legal states as if they were one. There is software that is not a device at all (Congress carved it out of the definition). There is software that is a device but that the FDA has cleared or approved. And there is software that may technically meet the device definition but that the agency states it will not actively regulate for now. That last state is enforcement discretion.

It is best read as a posture, not a permit. The FDA is saying it does not intend to enforce certain requirements at this time. It is not certifying that a product is safe, effective, or exempt, and it can revisit that stance later. Keeping the three apart is the most useful habit you can bring to the 2026 changes.

The CDS rules: four gates that all still stand#

The 2016 Cures Act added a provision that removes certain clinician-facing software from the device definition when it clears four gates at the same time. As the FDA restates them, non-device CDS software:

1. does not acquire, process, or analyze medical images, in-vitro-diagnostic signals, or patterns from a signal acquisition system;
2. displays, analyzes, or prints medical information about a patient or from peer-reviewed and clinical sources;
3. is meant to support or offer recommendations to a health professional about prevention, diagnosis, or treatment; and
4. lets that professional review the basis for the recommendation on their own rather than lean on it as the answer.

All four gates remain, and a tool has to pass all four. What the January revision altered is the interpretation layered over gate four.

The headline shift: one answer can now be enough#

Under the prior 2022 guidance, a tool that steered a clinician toward a single output, instead of laying out options to weigh, usually landed on the device side. The logic was that one answer left little room to satisfy gate four, so the software was effectively deciding rather than supporting.

The 2026 guidance opens a narrow enforcement-discretion policy for precisely that case. Per the FDA's document and contemporaneous legal analysis, the agency now says it will exercise enforcement discretion for CDS that yields a single output where only one recommendation is clinically appropriate, as long as the other non-device criteria hold. The stock example is software that flags a specific FDA-approved drug for a clinician to consider given a patient's symptoms and history. The catch is that the guidance never spells out how a developer should decide when a single recommendation is "clinically appropriate," which leaves the edge of the carve-out genuinely blurry and is where most of the practical disagreement will live.

Two more restrictions loosen#

Beyond the single-answer change, two older limits eased.

Time-critical decisions are no longer disqualifying on their own. Earlier guidance treated software meant for urgent, time-sensitive calls as outside non-device status almost by default. The 2026 version relocates that discussion so urgency is judged through the same lens as everything else: can a clinician meaningfully review the basis for the output?

Risk scores are back in play. The agency removed language that had singled out risk-probability or risk-score functions as inherently device-like. Those outputs are no longer flagged automatically.

The revision also touches documentation tools. Software that reviews a clinician's own findings to draft a proposed report and an associated recommendation can fall under enforcement discretion when a professional stays in the loop, provided the tool does not analyze the underlying image itself and draws only on well-accepted sources such as clinical guidelines and peer-reviewed literature.

Two boundaries are worth underlining. The guidance stays trained on clinician-facing software; it does not reach consumer symptom checkers, health chatbots, or patient-facing decision tools. And although the update was discussed publicly in terms of artificial intelligence, the text itself sets no AI-specific criteria.

The wellness side: what your watch can estimate#

The companion document governs a different world: products that make general wellness claims and carry low risk, which the FDA has long declined to regulate as devices. The core definition did not change. A product still qualifies only when it is intended for general wellness use and poses low risk.

What grew is the space for wearables that estimate physiologic parameters. The 2026 version signals that noninvasive products inferring measures such as heart rate variability, oxygen saturation, blood glucose, or blood pressure may be treated as general wellness products when several conditions all hold. The product must be noninvasive and avoid technologies that raise safety concerns, such as lasers, implants, or radiation. It must not be intended to diagnose, treat, cure, mitigate, or prevent disease. It must not substitute for an FDA-cleared device. It must avoid outputs that direct a specific clinical action, and it must not show clinical-style values unless those values are validated. Permitted outputs are ranges, trends, baselines, and longitudinal summaries across your sleep, activity, stress, and recovery. Anything aimed at diagnosis or disease management stays outside the category.

Reading the two documents as one move#

Put side by side, both revisions push in the same direction. More connected-device and software functionality is treated as non-device or covered by enforcement discretion, and the weight shifts toward how a product is designed, claimed, and documented rather than what it happens to measure. The trade-offs cut both ways. A broader discretionary zone can lower the compliance hurdle for tools that keep a clinician or a clear wellness framing in place, while the missing definitions for terms like "clinically appropriate" leave developers to guess where the line falls. Because guidance documents are nonbinding and enforcement discretion can be pulled back, the real reach of these changes will depend on how the FDA applies them over the coming years.