AI dental x-ray software reads bitewing, periapical or panoramic images and marks areas that may show decay or other findings, so the dentist can take a second look. Before you buy, check four things: what the FDA cleared it to do, how it was tested, how it fits your imaging and practice systems, and who controls the patient data.
This guide is for dental practice owners, group practice operators and dental service organizations comparing AI image tools. It covers the questions we ask before connecting any clinical tool to a practice's systems.
What does AI dental x-ray software actually do?
Most products in this category take an x-ray from your imaging software, run it through a trained model and return the image with highlighted areas. Common findings include suspected caries, bone level measurements, existing restorations and periapical radiolucencies. Some tools also add charting suggestions or patient-friendly images to help explain treatment.
The key word is aid. The software does not diagnose. The dentist still reviews every image and makes the call. Overjet's FDA clearance for its caries product, for example, states the tool is not intended as a replacement for a complete dentist's review
. That framing should shape how you train staff and how you talk about the tool with patients.
Is the software FDA cleared for your type of images?
Ask for the 510(k) number and read the clearance yourself on the FDA database. The indications for use tell you exactly what the tool may claim. Check three details:
- Image types. Overjet Caries Assist (K222746), cleared on March 27, 2023, covers bitewing and periapical radiographs. A tool cleared for bitewings is not automatically cleared for panoramic images.
- Findings. A clearance for caries detection does not cover bone loss or other findings unless the indications say so.
- Classification. Overjet's caries tool sits under 21 CFR 892.2070, Medical Image Analyzer, as a Class II device. Knowing the regulation helps you compare products on equal terms.
Clearances also change over time. Pearl announced on December 18, 2025 that it received FDA clearance to flag suspected caries, periapical radiolucencies and impacted third molars on panoramic x-rays. Ask each vendor which clearances cover the images your practice actually takes, and get that list in writing.
How was the AI tested, and on whose patients?
A model trained and tested on one population can perform differently on yours. Different sensors, exposure settings and patient mixes all matter. The American Dental Association has published guidance on this. ADA Technical Report No. 1109 explains how clinicians can evaluate AI image analysis systems, and ADA Standard No. 1110-1 sets requirements for the datasets used to validate them, including patient permissions and how the test data is held.
When the ADA released the proposed Technical Report 1109 in 2024, it highlighted the need for an independent dataset held by a party that does not make AI products, so accuracy can be compared fairly across tools. Ask vendors:
- How many images were in the test set, and were they separate from the training data?
- Which sensors and imaging systems were represented?
- What were the sensitivity and specificity for each finding, not just an overall score?
- Has anyone outside the company validated the results?
Will it fit your imaging and practice management systems?
This is where many rollouts stall. The AI needs a reliable path to receive images and a clear place to send results. Map the flow before you sign:
- Which imaging software and sensors do you use in each office, and does the vendor support all of them?
- Does the tool read images automatically, or does someone have to export them?
- Where do findings appear: inside the imaging viewer, in the practice management system, or only in a separate web portal?
- Can findings flow into treatment plans or reports without re-typing?
If results live only in a separate portal, your team will skip it on busy days. A small integration that puts findings where the dentist already works is often worth more than an extra feature. Our API integration services page explains how we approach these connections.
Who controls the patient data and the findings?
X-rays tied to a patient are protected health information. A cloud vendor that stores or processes them on your behalf is a business associate under HIPAA. HHS guidance says this applies even when the provider stores only encrypted data and has no key. So you need a signed business associate agreement before any images leave your systems.
Also ask:
- Will our images be used to train the vendor's models? Can we opt out?
- How long are images and findings kept after we cancel?
- Can we export all findings in a usable format if we switch tools?
- Do insurers or other third parties see any of this data through the vendor?
How should a practice run a pilot?
Start with one office and a set period, such as 60 days. Decide what you will measure before you turn it on:
- How often dentists agree or disagree with the flagged areas.
- Time spent per exam on image review.
- Whether treatment acceptance changes when patients see marked images.
- Any workflow steps the tool added for assistants or front desk staff.
Keep a simple log of disagreements. If the tool flags many areas the dentist rejects, staff will stop trusting it. If it misses findings the dentist catches, you need to know that before a wider rollout.
What mistakes should practices avoid?
- Assuming every image type is covered by the clearance.
- Treating AI marks as a diagnosis in patient conversations or notes.
- Skipping the business associate agreement for a free trial.
- Leaving results in a portal nobody opens.
- Rolling out to every office before one pilot has clear results.
If you are also reviewing other patient-facing tools, our guide on what clinics should check in telehealth software uses a similar checklist.
Frequently asked questions
No. Cleared tools are aids. They highlight areas for review, and the dentist makes the diagnosis and treatment decision.
Ask the vendor for the 510(k) number, then look it up in the FDA 510(k) database and read the indications for use, including which image types and findings are covered.
Yes, if the vendor stores or processes patient x-rays for you. HHS treats cloud providers handling protected health information as business associates, even when the data is encrypted.
Often, but support varies by vendor and system. Confirm which imaging and practice management systems are supported and where findings will appear before you sign.
Connecting AI tools to your practice systems?
We can map the image and data flow, build the integration and set up the reporting you need to judge a pilot.
Explore API integration servicesSources
- FDA: 510(k) clearance letter and indications for Overjet Caries Assist (K222746)
- Pearl: FDA clearance for panoramic x-ray AI (December 18, 2025)
- ADA: Standards on artificial and augmented intelligence in dentistry
- ADA News: Technical report calls for external validation of AI algorithms
- HHS: Guidance on HIPAA and cloud computing

