Questions to ask before buying a dental AI tool
Dental AI vendors increasingly approach clinics and labs. These questions help you separate solid products from marketing. They are based on published guidance for AI in dental research and on WHO's guidance on AI in health [1][2].
1. Is it a regulated medical device – where, and for what exactly?
- In the US, ask for the FDA 510(k) or De Novo number and look it up yourself in the FDA database [3]. In the EU, ask for the CE certificate, the device class and the notified body; the EU's medical device database is EUDAMED [4].
- Read the intended use / indications for use. Clearance applies to specific image types, findings and age groups. A product cleared for caries on bitewings is not automatically cleared for panoramic or CBCT images.
- A clearance in one country is not an authorisation in another.
2. What is the evidence?
- Was the tool tested on data from clinics, devices and populations similar to yours, not only on the developer's own data?
- Which reference standard was used (e.g. expert panel), and are sensitivity and specificity reported, not just “accuracy”? A consensus checklist for dental AI studies (31 items) covers how the reference test was built, comparison against relevant alternatives and performance metrics, and warns that plain accuracy can be misleading in imbalanced datasets [1].
- Is there any peer-reviewed evaluation, ideally independent of the company?
3. Data protection and security
- Where are patient images processed and stored (country, cloud provider)? In Finland and the EU, processing must comply with the GDPR.
- Is data used to retrain the vendor's models, and can you opt out?
4. Workflow, cost and responsibility
- Does it integrate with your imaging and practice-management software?
- How are updates handled, and do model updates change the regulatory status?
- Who is responsible when the AI misses something or flags a false positive? WHO stresses human oversight, transparency and accountability for AI in health [2].
- Will it be used to explain findings to patients? Decide in advance how you avoid over-treatment.
Tip: our AI tools directory links each regulatory claim to its official source, so you can check it yourself.
Sources
- Schwendicke F, et al. Artificial intelligence in dental research: Checklist for authors, reviewers, readers. J Dent. 2021;107:103610. https://doi.org/10.1016/j.jdent.2021.103610
- World Health Organization. Ethics and governance of artificial intelligence for health. 2021. https://www.who.int/publications/i/item/9789240029200
- US FDA. Premarket Notification 510(k) database. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm
- European Commission. EUDAMED – European database on medical devices. https://ec.europa.eu/tools/eudamed