FDA regulatory foundations
What Is a Medical Device?
CAHIR Solutions · Last reviewed: August 2026
The FDA starting point: intended use
Under section 201(h) of the FD&C Act, a device is an instrument, apparatus, machine, implant, in vitro reagent or similar article — including software — intended for use in the diagnosis, cure, mitigation, treatment or prevention of disease, or intended to affect the structure or any function of the body, without achieving its primary purpose through chemical action or metabolism.
The operative word is intended. FDA reads intent from your labeling, instructions, marketing copy, sales conversations, website claims and even the wording on your pricing page. Two products with identical code can land in different regulatory worlds because one says "track your resting heart rate" and the other says "detect irregular rhythms." Write the intended-use statement before you write anything else, and keep every public claim consistent with it.
Screening outputs on this page are signals for planning, not determinations. Only FDA can classify a product.
Five-step medical-device screen
Work through these steps in order. Each one narrows the question, and a single step can be decisive on its own.
01
Write the intended use
State in one sentence what the product is for, for whom, and in what setting. Everything downstream follows from that sentence, including labeling, evidence and class.
A stated medical purpose pushes toward device.
02
Check for a disease or condition claim
Does the product reference diagnosis, cure, mitigation, treatment or prevention of a disease, or affect the structure or function of the body? General wellness claims sit outside; condition-specific claims do not.
Naming a condition pushes toward device.
03
Characterise the output
Distinguish raw data display from an interpretation. Logging a heart rate is different from flagging a possible arrhythmia, and a ranked differential is different from a specific directive to act.
Interpretation and directives push toward device.
04
Identify the user and the setting
A tool for a clinician making a time-critical call in an acute setting carries a different risk profile from a reference tool a consumer reads at home with time to reflect.
Time-critical clinical use pushes toward device.
05
Test the software-function exclusions
Check the administrative, records, general wellness, and clinical decision support exclusions in section 520(o), as well as medical device data system status where the product only transfers, stores, converts or displays data.
Failing any exclusion criterion pushes toward device.
A product that clears all five steps is a candidate for non-device status, not a confirmed non-device.
When a non-device becomes a device
Most teams do not set out to build a regulated product. They cross the line by adding a feature or sharpening a marketing claim. These pairs show the most common crossings.
Activity tracker that displays heart rate trends
↓ Same tracker that notifies the user of possible atrial fibrillation
The output changed from data display to a condition-specific interpretation.
Scheduling and billing tool that sequences appointments
↓ Same tool that prioritises patients by predicted deterioration risk
The product now drives clinical management rather than administration.
General health chatbot that summarises public guidance
↓ Chatbot that takes symptoms and recommends a specific treatment
A specific preventive, diagnostic or treatment output defeats the decision-support exclusion.
Image viewer that renders DICOM studies for review
↓ Viewer that segments and flags suspicious lesions automatically
Analysing an image from a device removes the software from the exclusion entirely.
Feature roadmaps change regulatory status. Re-screen whenever claims or outputs change.
Digital-health and software examples
The buckets below reflect how comparable products have typically been treated. They are orientation aids — your specific intended use governs.
Generally not a device
- Practice-management, scheduling and claims software
- Electronic health records used for storage and display
- General wellness apps: sleep, fitness, mindfulness, healthy weight
- Medical reference libraries and clinical education content
- Software that only transfers, stores, converts or displays device data
Likely a device
- Computer-aided detection or triage on imaging studies
- Algorithms interpreting ECG, EEG or other physiological signals
- Dosing calculators that output a specific therapeutic directive
- Continuous monitoring that alarms on a clinical threshold
- Digital therapeutics claiming to treat a diagnosed condition
Needs further analysis
- Risk-prediction models embedded in clinician workflow
- Symptom checkers and triage assistants for consumers
- Large-language-model assistants that summarise patient charts
- Remote-monitoring platforms that combine display with alerting
- Wellness products whose marketing copy drifts toward condition claims
Placement here is illustrative only and is not an FDA classification of any named product category.
FDA tools for deeper research
Once you have a candidate intended use, these primary sources let you test it against how FDA has actually classified comparable products.
Database results are evidence for a screening view, not a classification decision.
Frequently asked questions
Is my wellness app a medical device?
Usually not, if it makes only general wellness claims — encouraging exercise, sleep, relaxation or healthy weight — and does not reference a disease or condition. FDA's general wellness policy turns on the claim, not the sensor. Once the app says it detects, diagnoses, monitors or helps treat a specific condition, it moves into device territory.
How do I know whether an FDA-cleared predicate already exists?
Search the FDA Product Classification Database for a regulation description that matches your intended use, note the three-letter product code, then search the 510(k) and De Novo databases for that code. Existing clearances with similar intended-use statements are a strong signal that a 510(k) route is available.
Is clinical decision support software regulated?
Some clinical decision support is excluded from the device definition when it displays or analyses medical information, offers a recommendation rather than a specific directive, does not analyse a signal or image from a device, and lets the clinician independently review the basis of that recommendation. Time-critical outputs, signal or image analysis, and opaque model reasoning each defeat the exclusion.
Does using AI or machine learning change the answer?
AI does not by itself make a product a device, and it does not exempt one either. The intended use still governs. What AI changes is the evidence and lifecycle burden: performance across subgroups, transparency of the recommendation basis, and a plan for how retraining and model updates are controlled after authorization.
Do I need clearance before I can sell?
If your product meets the device definition and is not exempt, marketing it without the required clearance, De Novo authorization or approval is a violation. Exempt Class I and some Class II devices can be marketed after registration and listing without a premarket submission, but they still carry quality-system and labeling obligations.
What about the EU and UK?
The EU MDR and the UK regime use their own definitions and classification rules, and they frequently classify software higher than FDA does. A product that is an unregulated wellness tool in the US can be a Class IIa device in the EU under MDR Rule 11, so screen both jurisdictions in parallel rather than sequentially.
Continue your regulatory planning: Read the SaMD regulatory guide
This guide is informational only and is not legal or regulatory advice. Confirm any classification decision against current FDA guidance and, where needed, with qualified regulatory counsel.