AI in dentistry: what the studies and FDA clearances show
Only a paper we could open, and only FDA records we could open. No accuracy figure is added from a paper that does not state one.
Short answer
Last checked against the case studies linked on this page. Those cases keep their own source dates.
The review we opened, PMC11426822 (DOI 10.4103/jpbs.jpbs_1290_23), describes applications of AI in dentistry and names limits. It does not report a patient sample, an accuracy percentage, or a time or money result for a practice. The FDA table is a clearance list we checked, not a ranking of products.
The study we opened
Revolutionizing Dentistry: The Applications of Artificial Intelligence in Dental Health Care. Journal of Pharmacy and Bioallied Sciences. Narrative review. Opened 23 October 2026 at PMC11426822. The text lists uses such as diagnostic imaging, treatment planning, and administrative tasks, and it discusses limits including data, privacy, and fitting the tools into a clinic. It does not give a number of patients studied, a sensitivity, or hours saved in a named practice. The article page states that financial support was nil and that the authors declared no conflicts. We do not repeat the correspondence address or email.
A systematic review at PMC7770315 was not used. The fetch returned 403, so none of its numbers appear here.
FDA clearances we opened
These records are the same set as on the Pearl and Overjet comparison. Practice results, where we have them, stay on the case pages: Blue Court Dental for Pearl's imaging case, and Bridge Mill Dental Care for Overjet's insurance-verification case.
| Company | Device | 510(k) | Decision date | Decision | What we opened | What that source says it detects | Link |
|---|---|---|---|---|---|---|---|
| Pearl | Second Opinion | K210365 | Substantially equivalent | Decision letter | The letter says it is computer-aided detection software for bitewing and periapical radiographs of permanent teeth in patients 12 or older, used as a second reader. It names caries, margin discrepancy, calculus, periapical radiolucency, and restorations (crowns, bridges, implants, root canals, fillings). The letter says an MRMC study showed statistically significant aided-reader improvement for those four pathologies. The letter text we used does not state a sensitivity or specificity figure. | FDA record | |
| Pearl | Second Opinion Panoramic | K250525 | Substantially equivalent | Database record only | Not stated here. The database record names the device, the decision date, substantial equivalence, and product code MYN. The indications PDF was not opened. | FDA record | |
| Pearl | Second Opinion Periapical Radiolucency Contours | K242600 | Substantially equivalent | Database record only | Not stated here. The database record names the device, the decision date, substantial equivalence, and product code MYN. The indications PDF was not opened. | FDA record | |
| Overjet | Caries Assist | K212519 | Substantially equivalent | Database record only | Not stated here. The database record names the device Overjet Caries Assist, the decision date, substantial equivalence, and product code MYN. The indications PDF was not opened. | FDA record | |
| Overjet | Caries Assist | K222746 | Substantially equivalent | Database record only | Not stated here. The database record names the device, the decision date, and substantial equivalence. The indications PDF was not opened. | FDA record | |
| Overjet | Periapical Radiolucency Assist | K231678 | Substantially equivalent | Decision letter | The letter says it aids detection of periapical radiolucencies on permanent teeth on periapical radiographs, as a concurrent read, for patients 12 or older. It is not a replacement for the dentist, and it is for prescription use. | FDA record | |
| Overjet | Caries Assist-Pediatric | K233738 | Substantially equivalent | Database record only | Not stated here. The database record names the device, the decision date, and substantial equivalence. The indications PDF was not opened. | FDA record | |
| Overjet | Multi-Image Caries and Charting Assist | K261059 | Substantially equivalent | Database record only | Not stated here. The database record names the device, the decision date, substantial equivalence, product codes MYN and QIH, and that a predetermined change control plan was authorized. The indications PDF was not opened. | FDA record |
These are the 510(k) records opened on 23 October 2026. This is not a complete list of either company's clearances. A 510(k) decision means the FDA found the device substantially equivalent to a legally marketed predicate. It is not an FDA approval. Opened .
What studies don't tell a practice owner
A narrative review can list what AI is used for. It does not tell you the price a vendor will charge your office, whether the software writes back to your practice management system, or whether a vendor's hours-saved claim was measured. A 510(k) letter can say what a device is indicated to detect, and for whom. It is not a before-and-after at a named practice. Where this site has a practice result, it is on a case page, with a source label. Vendor-reported numbers stay labeled as the vendor's claims.
The Blue Court case's 73% figure is Pearl's label for 32 of 44 lesions, not a result copied from PMC11426822. The Bridge Mill hours are Overjet's claims about insurance checks, not a finding of the review.
Questions
What does the study we opened show?
We opened Revolutionizing Dentistry: The Applications of Artificial Intelligence in Dental Health Care, a narrative review in the Journal of Pharmacy and Bioallied Sciences (DOI 10.4103/jpbs.jpbs_1290_23, PMC11426822), on 23 October 2026. It lists applications, including diagnostic imaging, and it lists limits. The text we read does not report an original patient sample, an accuracy percentage, or a time or money result for a practice. The page states that financial support was nil and that the authors declared no conflicts. Correspondence details are not repeated here.
Which FDA clearances are on this page?
Only the eight records opened on 23 October 2026 and listed in the table. Two decision letters were read: K210365 and K231678. The other rows are database records, and their indications PDFs were not opened. The list is not complete. A 510(k) decision is substantial equivalence, not an FDA approval.
What do studies not tell a practice owner?
The review we opened does not say what a named practice paid, which system it runs, or whether a vendor's hours-saved claim was measured. A clearance is not a result in your office. The Blue Court and Bridge Mill cases are vendor reports about two different jobs, and this page does not turn the review into a reason to buy either product.
Why is only one study cited?
A study is cited here only if its full text was opened. A fetch of a systematic review at PMC7770315 returned 403, so its numbers are not used. Third-party prices and performance percentages that were not in an opened 510(k) letter are left out.
The same text is at /research.md.