What FDA Clearance Means for AI Medical Software (and How to Check Any Device)
FDA cleared, FDA approved and FDA registered mean very different things. Here's what each label says about AI medical software and how to look up any device yourself.
FDA clearance means the FDA found an AI medical device substantially equivalent to a legally marketed device through the 510(k) pathway. It is not the same as FDA approval, which applies to high-risk devices through premarket approval. Novel devices may instead be granted De Novo authorization. You can confirm any device's status in FDA's public databases.
Photo: Unsplash
On this page
- Cleared, approved, granted: the short version
- What 510(k) clearance actually tells you
- Which AI health software needs FDA review at all
- How FDA handles AI that changes over time
- How to look up any AI device in FDA’s databases
- Red flags in AI health marketing
- Where to start if you’re evaluating an AI tool
- Conclusion
- Sources
A vendor tells you its AI tool is “FDA cleared.” A health app says it’s “FDA registered.” A news story calls a new algorithm “FDA approved.” Are these the same thing? And does any of it mean the software actually works for your patients?
These labels have specific legal meanings, and mixing them up is one of the most common mistakes clinics, buyers and patients make when evaluating AI in healthcare. This guide explains what FDA clearance means for AI medical software, how it differs from approval and other pathways, and how to look up any device in FDA’s public databases in a few minutes.
Cleared, approved, granted: the short version
FDA regulates medical devices by risk, and software can be a medical device. The word used for a device’s status depends on the pathway it went through.
| Label | Pathway | Typical risk level | What FDA decided |
|---|---|---|---|
| Cleared | 510(k) premarket notification | Low to moderate (Class I or II) | The device is substantially equivalent to a legally marketed “predicate” device |
| Granted | De Novo classification request | Novel, low to moderate risk with no predicate | A new device type was classified, with controls; it can then serve as a predicate |
| Approved | Premarket approval (PMA) | High risk (Class III) | The maker independently showed reasonable assurance of safety and effectiveness |
| Exempt | No premarket review | Lowest risk | General controls apply, but FDA did not review it before sale |
| “FDA registered” | Establishment registration and listing | Any | Nothing about safety or performance; FDA says registration does not mean approval, clearance or authorization |
The key point: “FDA cleared” is not “FDA approved.” Under 21 CFR 807.97, a 510(k) clearance does not constitute FDA approval, and representing it as official approval is misleading and counts as misbranding.
What 510(k) clearance actually tells you
FDA describes the 510(k) standard as comparative. The maker shows that its device is as safe and effective as, or “substantially equivalent” to, a predicate device already on the market. For AI software, that usually means comparing the algorithm’s intended use, technology and performance data to an earlier device.
Clearance tells you that:
- FDA reviewed the device for a specific intended use, set out in its Indications for Use statement
- The maker submitted data showing substantial equivalence for that use
- The device can be legally marketed in the US for that use
Clearance does not tell you:
- That the tool will perform the same way on your patient population, scanners or workflow
- That it’s better than other tools or than clinicians working without it
- That uses outside the cleared indication are covered
That last point matters. If an AI tool is cleared to flag suspected findings on chest CT for triage, it isn’t cleared to diagnose those findings on its own, or to read other kinds of images.
Key takeaway: Clearance is permission to market a device for a defined use, based on comparison to an existing device. Always read the Indications for Use, because that’s the boundary of what FDA actually reviewed.
Which AI health software needs FDA review at all
Not every AI tool in healthcare is a regulated medical device. FDA’s oversight depends on what the software is intended to do.
Usually regulated as a device:
- Software that analyzes medical images, signals or lab data to detect, diagnose, triage or monitor disease
- Algorithms that drive treatment, such as dosing recommendations without room for clinician review
- Software that provides diagnostic or treatment recommendations directly to patients or caregivers
Often outside device oversight:
- Administrative tools such as scheduling, billing and coding software
- Certain clinical decision support (CDS) software for health care professionals that meets criteria in the FD&C Act, including letting the clinician independently review the basis for its recommendations. FDA re-issued its final CDS guidance in January 2026.
- Low-risk general wellness products. FDA reissued its General Wellness guidance on January 6, 2026, clarifying when non-invasive sensing products fall in this category.
Many ambient AI scribes and documentation tools fall into the administrative or non-device category, which is why they rarely carry FDA clearance. See ambient AI scribes in the doctor’s office for how clinics evaluate them instead. Whether a specific product is a device depends on its exact intended use and claims, so check with the vendor and FDA’s guidance. This is not legal advice.
How FDA handles AI that changes over time
Traditional devices are “locked”: the version FDA reviewed is the version sold. AI models can be retrained, which raises a question about when an update needs a new submission.
FDA’s answer is the predetermined change control plan (PCCP). FDA issued final guidance in December 2024 on PCCPs for AI-enabled device software functions. A PCCP, reviewed as part of the original submission, describes planned modifications, how they’ll be developed and validated, and how their impact will be assessed. Changes made within an authorized PCCP can be implemented without a new marketing submission. Changes outside it generally need one.
If a vendor says its model “learns” or updates, ask whether the device has an authorized PCCP and what changes it covers.
How to look up any AI device in FDA’s databases
You don’t have to take a vendor’s word for it. Here’s how to check in about 5 minutes.
Step 1: Get the submission number
Ask the vendor for the FDA submission number. It looks like:
- K followed by 6 digits (for example, K123456) for a 510(k) clearance
- DEN followed by 6 digits for a De Novo grant
- P followed by 6 digits for a PMA approval
A legitimate vendor will give you this readily. It’s often in the product’s labeling or regulatory page.
Step 2: Check FDA’s AI-enabled device list
FDA maintains a public List of Artificial Intelligence-Enabled Medical Devices, showing each device’s decision date, submission number, device name, company, lead review panel and product code. The page was current as of September 4, 2026, and entries date back to 1995. You can sort it or download it as CSV, Excel or XML. FDA notes the list is not comprehensive, because devices were identified mainly by AI-related terms in their public summaries. If a device isn’t on it, check the main databases next.
Step 3: Search the main databases
- Releasable 510(k) database: search by 510(k) number, device name, product code or panel. FDA updates it weekly.
- De Novo database: search by device name or requester (company) name.
- PMA database: for high-risk devices.
- Devices@FDA: a combined catalog of cleared and approved device information, searchable by device or company name.
Step 4: Read the decision documents
Open the 510(k) Summary or De Novo decision summary. Look for:
- Indications for Use: the exact use FDA reviewed
- Predicate device: what it was compared to
- Performance testing: study design, sample size, sensitivity and specificity, and the populations and equipment tested
- Limitations and warnings
Step 5: Match the company and product name
Confirm the company name and device name match what you’re buying. Some vendors sell several products, and only one may be cleared.
Red flags in AI health marketing
| Claim | What to check |
|---|---|
| “FDA registered” or “FDA certified” | FDA doesn’t issue registration certificates. Registration says nothing about safety or performance |
| “FDA approved” for a 510(k) device | Inaccurate. Ask for the K number and check the status |
| “Cleared for diagnosis” | Read the Indications for Use; many tools are cleared for triage or decision support only |
| “Our AI learns continuously” | Ask whether there’s an authorized PCCP covering those changes |
| No submission number available | Treat as unreviewed until proven otherwise |
| Clearance for an older product version | Confirm the version you’re buying is the one covered |
Where to start if you’re evaluating an AI tool
- Ask for the submission number and verify it in FDA’s databases.
- Read the Indications for Use and compare it to how you plan to use the tool.
- Review the performance data in the decision summary, including who was in the test population.
- Ask for real-world evidence from settings like yours, and run a local pilot before full rollout.
- Check privacy and security separately. FDA clearance doesn’t cover HIPAA compliance. See is your health app covered by HIPAA.
- Plan for monitoring. Track performance after go-live, and report problems through FDA’s MedWatch program.
Patients evaluating AI-read results can start with AI in radiology: what patients should know. For consumer devices, see are smartwatch ECG readings accurate and can a phone app check a mole. The bigger picture is in AI in healthcare 2026.
Conclusion
“FDA cleared” means a device was found substantially equivalent to an existing device for a specific use. It’s a meaningful review, but it’s not FDA approval and it doesn’t guarantee performance in your setting. The good news is that FDA’s databases are public, free and quick to search. Ask for the submission number, read the Indications for Use and the decision summary, and treat “FDA registered” as a marketing phrase, not a safety signal.
This is general information, not medical advice. This is not legal advice.
Sources
- FDA: Medical device safety and the 510(k) clearance process — 510(k) as a comparative standard, predicates, risk-based pathways
- FDA: Premarket submissions: selecting and preparing the correct submission — 510(k), De Novo, PMA and HDE pathways
- eCFR: 21 CFR Part 807 Subpart E (including 807.97) — clearance does not constitute approval
- FDA: List of AI-enabled medical devices — list contents, columns, downloads and scope limits
- FDA: Are there “FDA registered” or “FDA certified” medical devices? — registration vs clearance, Devices@FDA and De Novo database
- FDA: Guidances with digital health content and FDA Roundup, December 3, 2024 — 2026 CDS and General Wellness guidances, PCCP final guidance
Frequently asked questions
Is FDA cleared the same as FDA approved?
No. "Cleared" means FDA found a device substantially equivalent to a legally marketed predicate device through the 510(k) pathway, typically for low- to moderate-risk devices. "Approved" applies to high-risk Class III devices that went through premarket approval, which requires independent evidence of safety and effectiveness. Federal regulations say a 510(k) clearance does not constitute FDA approval.
How do I check if an AI medical device is FDA cleared?
Ask the vendor for its submission number, such as a K number for a 510(k). Then search FDA's releasable 510(k) database, De Novo database or Devices@FDA by number, device name or company. You can also check FDA's List of AI-Enabled Medical Devices, though FDA notes that list isn't comprehensive. Read the Indications for Use in the decision summary.
What does "FDA registered" mean for a medical device?
It only means the company registered its facility and listed its device with FDA, which is generally required for device makers and distributors. FDA says registration and listing do not mean a device is approved, cleared or authorized. FDA also doesn't issue registration certificates, so a "certificate" shown to imply FDA review is misleading.
Does FDA clearance mean an AI tool is accurate?
Not on its own. Clearance means FDA reviewed data showing the device is substantially equivalent to a predicate for a specific intended use. Accuracy can differ in your patient population, equipment or workflow. Read the performance testing in the 510(k) Summary, ask for real-world evidence from similar settings and run a local pilot.
Do all AI health apps need FDA clearance?
No. Administrative tools for scheduling, billing and documentation usually aren't medical devices. Some clinical decision support software for clinicians and low-risk general wellness products also fall outside FDA device oversight under FDA guidance updated in January 2026. Software that diagnoses, triages or treats disease generally does need FDA review. This is not legal advice.
What is a De Novo authorization?
De Novo is FDA's pathway for novel, low- to moderate-risk devices that have no predicate to compare against. If FDA grants the request, it creates a new device classification with required controls, and the device can then serve as a predicate for future 510(k) submissions. Many first-of-their-kind AI tools reach the market this way.
Can FDA cleared AI software be updated after clearance?
Yes, within limits. Changes covered by an authorized predetermined change control plan (PCCP) can be made without a new submission, as described in FDA's final guidance on PCCPs for AI-enabled device software functions. Significant changes outside an authorized plan generally require a new marketing submission. Ask vendors whether their device has a PCCP.
Was this guide useful
No reviews yet. Share your experience and help other readers.
Write a review
Manishdai
Manishdai writes about the technology that is reshaping every industry — from AI agents and cloud platforms to healthtech, fintech and the IT companies building it all across the United States.
manishdai.com