What Changed in FDA's 2026 Clinical Decision Support Guidance

On January 6, 2026, the U.S. Food and Drug Administration issued revised final versions of two long-standing digital health guidance documents: one on Clinical Decision Support (CDS) software and one titled General Wellness: Policy for Low Risk Devices. The CDS revision supersedes the 2022 guidance, and the general wellness revision supersedes the 2019 version. In plain terms, both documents widen the set of software and connected-device functions that the agency treats as outside active device regulation, either because they are not devices at all or because the FDA says it will exercise enforcement discretion. Neither document changes the underlying statute; each changes how the FDA reads it.

This article explains the mechanics of what shifted and which software now sits in the discretionary zone. It is educational and not medical or regulatory advice.

The statutory frame that has not changed

The dividing line for CDS software comes from Section 520(o)(1)(E) of the Federal Food, Drug, and Cosmetic Act, added by the 2016 Cures Act. That provision carves certain software out of the definition of a medical device when it meets four criteria at once. As restated in the FDA's guidance, non-device CDS software (1) does not acquire, process, or analyze medical images, signals from an in vitro diagnostic device, 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 intended to support or provide recommendations to a health care professional about prevention, diagnosis, or treatment; and (4) enables that professional to independently review the basis for the recommendation rather than rely primarily on it.

All four still apply, and all four must be met. What the 2026 revision changed is the interpretation layered on top of them.

The central CDS change: a single recommendation can now qualify

Under the 2022 guidance, software that pointed a clinician toward one specific output, rather than presenting a list of options to weigh, tended to fall on the device side of the line. The reasoning was that a single answer left little room for independent evaluation under Criterion 4, so the tool was doing more than supporting a decision.

The 2026 guidance introduces a limited enforcement discretion policy for exactly that situation. According to the FDA's document and contemporaneous legal analysis, the agency now says it will exercise enforcement discretion for CDS tools that provide a singular output where only one recommendation is clinically appropriate, provided the other non-device criteria are met. The commonly cited example is software that identifies a specific FDA-approved drug for a clinician to consider based on a patient's symptoms and history. That said, the guidance does not define how a developer should determine when a single recommendation is clinically appropriate, which leaves the boundary of the carve-out unsettled.

The legal category here deserves precision. Enforcement discretion is not approval, clearance, or a statement that the software is not a device. It is a stated intention by the agency not to enforce certain requirements at a given time, which the FDA can revisit.

Other CDS adjustments

Two further changes narrow older restrictions. First, the FDA relocated its treatment of time-critical decision-making. Earlier guidance effectively excluded software meant to inform urgent, time-sensitive decisions from non-device status; the 2026 version moves that discussion so that a time-critical setting is no longer an automatic disqualifier and is instead evaluated through whether a clinician can meaningfully review the basis for the output. Second, the agency removed prior language that had singled out risk probability or risk score functions as failing the criteria, so those functions are no longer flagged as inherently device-like.

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

Two limits are notable. The guidance remains focused on clinician-facing software and does not extend to consumer-facing symptom checkers, health chatbots, or patient decision tools. And although the update was framed publicly around artificial intelligence, the text itself does not set out AI-specific criteria.

The general wellness update

The companion guidance addresses a different category: products that make general wellness claims and pose low risk, which the FDA has long said it does not actively regulate as devices. The definition itself is unchanged. Products still qualify only if they are intended for general wellness use and present low risk to users.

What expanded is the room for wearables that estimate physiologic parameters. The 2026 version indicates that noninvasive products estimating or inferring parameters such as heart rate variability, oxygen saturation, blood glucose, or blood pressure may be treated as general wellness products when a set of conditions holds: the product is noninvasive and does not use a technology (such as lasers, implants, or radiation) that raises safety concerns; it is not intended to diagnose, treat, cure, mitigate, or prevent disease; it does not substitute for an FDA-cleared device; it avoids claims or outputs that direct specific clinical action; and it does not display clinical-style values unless those values are validated. Permitted outputs include ranges, trends, baselines, and longitudinal summaries in areas like sleep, activity, stress, and recovery. Products intended for diagnosis or disease management remain outside the category.

How to read the two documents together

Both revisions move in the same direction: more connected-device and software functionality treated as non-device or subject to enforcement discretion, with the burden shifting toward how a product is designed, claimed, and documented rather than what it measures. The trade-offs are real on both sides of the ledger. A wider discretionary zone can lower the compliance barrier for tools that keep a clinician or a wellness framing firmly in place, while the absence of defined tests for terms like clinically appropriate leaves developers to interpret where the line sits. Because guidance documents state nonbinding recommendations and enforcement discretion can be withdrawn, the practical scope of these changes will depend on how the FDA applies them over time.