Is AI in Dentistry Working for Everyone?

By Genni Burkhart, Editor

In the United States, dentists take nearly 100 million dental radiographs every year, and artificial intelligence (AI) now has a first look at those images, long before a human weighs in. That pace of AI integration into patient care is why Dr. Hawazin Elani, an assistant professor of oral health policy and epidemiology from Harvard School of Dental Medicine, offered this advice at the school's Global Symposium on AI in Dentistry, “The first rule is to do no harm. If we as a profession don't consider algorithmic bias and fairness, we will cause unintentional harm.”

For dentists, the issue extends beyond the technology itself. AI is a tool, and like any tool, its value depends on how it's used. Much of the data behind it is built around the average patient. Yet few patients fit neatly into an average. As these tools become more common in clinical practice, it becomes easy to assume their recommendations work equally well for everyone. Whether that assumption holds is a question dentistry is just beginning to explore.

Is AI Inheriting Human Bias?

Essentially, AI can only learn from the data it's fed. This means a dental AI model built on the average mouth performs best on, well, the average mouth. Anything outside of typical, and AI accuracy declines. When Dr. Elani reviewed hundreds of studies on AI in dentistry, she found only two that examined fairness or disparities, along with evidence that some tools produce more accurate diagnoses for white patients than for patients of color. However, the software didn't choose that gap. The people who built it and picked its training data did, even without the intention to do so.

That raises an important question. Who decides what "typical" looks like in the first place? Healthcare has a long history of building research, diagnostic criteria, and treatment recommendations around a relatively narrow patient population, then treating everyone outside that group as an exception. AI risks inheriting those same assumptions. If the training data doesn't adequately reflect the diversity of patients seen in practice, the technology may define "normal" far more narrowly than reality does.

A 2025 review of 66 studies on disability and algorithmic fairness found that developers routinely build these datasets without consideration for patients with disabilities. Those models then read atypical patients less accurately while appearing to have the same confidence in accuracy. Further compounding this issue, atypical patients face higher risks of being re-identified from supposedly "anonymous" health data. Meaning, patients most likely to be misread are also the most exposed.

The Authority Problem

The Disability Rights Education and Defense Fund describes another well-documented, problematic issue with AI. Once an algorithm offers an answer, people tend to defer to it even when it conflicts with their own expertise. And the higher the stakes, the higher the level of deference. The concern isn't that clinicians will stop making decisions altogether. It's that the role AI plays in those decisions can gradually become larger than intended. However it's used, AI is a tool, not a decision-maker. Because a computer has no concept of fairness, it cannot decide for itself what counts as such. It can only return patterns from the data it's programmed with. The inherent bias lies with the input and in the human choices behind the programming. The problem isn't that AI has too much authority; it's forgetting who bears responsibility for the final decision.

Bias Limits Access

Hawazin W. Elani, BDS, PhD, and William V. Giannobile, DDS, DMSc, writing in JAMA Health Forum, argue that dentistry has a rare chance to get this right.

Building new technology on an old, narrow picture of the typical patient prevents AI from reaching its full capabilities. With more than 75 million Americans living in dental professional shortage areas, AI-driven tools have a real opportunity to extend care to many of them. However, that promise holds only if the profession insists the technology works for every patient, not just the ones who are easiest to study. Whether AI helps reduce oral health disparities or reinforces them will depend largely on the choices made now.

Human Advantage

The BDJ Open researchers found that prior training in AI ranked among the strongest predictors of whether dentists used these tools at all, a topic explored in Part 1 of this series on AI adoption in dental education. Essentially, a dentist who understands AI as a tool reads its output with judgment, questions results that look off, and stops a "wrong answer" before it ever reaches the patient. Fortunately for patients, survey data shows dentists already treat AI this way, with most who use it calling it a helpful second opinion, and only a small fraction admitting to trusting it over their own clinical judgment.

AI touches nearly every aspect of modern life, from healthcare to entertainment, the economy, transportation, education, science, and research. As such, it deserves our full attention and warrants further understanding. The evidence behind its use in oral healthcare, though, still reflects that some patients are far better served than others. As discussed throughout this series, AI has enormous potential to improve patient care. However, a tool built on the "typical" patient will continue to serve a select population more accurately than a full spectrum of patients it was never trained to understand. Do no harm has always been the guiding principle of healthcare. As AI becomes more integrated into dentistry, that responsibility extends to the data, assumptions, and algorithms helping shape clinical decisions.

Author: With over 16 years as a published journalist, editor, and writer, Genni Burkhart's career has spanned politics, healthcare, law, business finance, technology, and news. She resides in Northern Colorado, where she works as the editor-in-chief of the Incisor at DOCS Education.

References

1. Elani HW, Giannobile WV. Harnessing artificial intelligence to address oral health disparities. JAMA Health Forum. 2024;5(4):e240642. https://doi.org/10.1001/jamahealthforum.2024.0642
2. McAlpine KJ. AI, dentistry, and ethics: prioritizing transparency and equitable access. Harvard School of Dental Medicine. June 10, 2024. https://hsdm.harvard.edu/news/ai-dentistry-and-ethics-prioritizing-tran…
3. Ramsey L. Artificial intelligence in patient care, education and treatment. Metro Denver Dental Society. April 2, 2025. https://mddsdentist.com/practice-management/ai-in-dentistry/
4. Pai M, Yellapurkar S, Chengappa SK, Pentapati KC. Exploring the prospects of artificial intelligence in transforming dental care for special needs groups: mapping the current evidence. BDJ Open. 2026;12(1):49. https://doi.org/10.1038/s41405-026-00436-x
5. Dall C. FDA clears first AI-based early warning system for sepsis. CIDRAP. 2026. https://www.cidrap.umn.edu/sepsis/fda-clears-first-ai-based-early-warni…
6. Vogt Y. Disability and algorithmic fairness in healthcare: a narrative review. J Med Artif Intell. 2025;8:56. https://doi.org/10.21037/jmai-24-415
7. Disability Rights Education and Defense Fund. Disability bias in clinical algorithms: recommendations for healthcare organizations. December 7, 2023. https://dredf.org/disability-bias-in-clinical-algorithms/
8. Portalatin A. 32% of dentists using AI: study. Becker's Dental Review. May 27, 2026. https://www.beckersdental.com/ai-teledentistry/40-of-dentists-quit-ai-t…