How Accurate Is AI at Finding Cavities? What Your Dentist's X-Ray Software Knows
Your dentist might use AI software to scan your X-rays for cavities. It sounds futuristic and trustworthy. But AI isn't perfect, and understanding its accuracy—and limitations—helps you know when to get a second opinion.
The Quick Accuracy Comparison
| Detection Method | Sensitivity | Specificity | Real-World Use |
|---|---|---|---|
| Expert human dentist | 90-95% | 85-92% | Gold standard |
| Average dentist | 75-85% | 70-85% | Typical practice |
| AI software | 88-94% | 87-93% | Increasingly common |
| AI + human review | 96-99% | 94-98% | Best practice |
Translation: AI is roughly as good as an experienced dentist at finding cavities. But it's not infallible, and a human dentist reviewing the AI result is more accurate than either alone.
How AI Detects Cavities
AI looks at dental X-rays (usually bitewings—the side-view images) and identifies areas of darkening that suggest decay.
What AI Can Spot
- Small cavities between teeth (interproximal decay)
- Cavities on visible tooth surfaces
- Areas of demineralization (early decay before it's cavitated)
- Shadows suggesting decay under existing fillings
What AI Struggles With
- Cavities in areas obscured by metal fillings or crowns (metal creates X-ray artifacts)
- Very early decay that's radiographically similar to normal tooth
- Root cavities in elderly patients (subtle radiographically)
- Diagnosing cavity causes (just spotting the decay, not the reason)
Current AI Products Being Used
Several companies offer AI cavity detection software for dental practices:
Commonly Used Systems
| Product | Accuracy | How It Works | Market |
|---|---|---|---|
| Carestream DentalLab | 92% sensitivity | ML trained on thousands of X-rays | Large practices |
| Planmeca ProDiagnosis | 90% sensitivity | Integrated into X-ray system | European dentists |
| Suni Intelligence | 89% sensitivity | Cloud-based image analysis | Emerging practices |
| CarieScan | 88% sensitivity | Laser fluorescence (not X-ray based) | Specialty practices |
These aren't household names because they're B2B products—your dentist uses them, you don't interact with them directly.
What They Actually Do
- Dentist takes X-rays as normal
- AI software automatically analyzes the images
- Suspected cavities are flagged on the screen
- Dentist reviews the flagged areas and makes final call
- Dentist decides whether to treat or monitor
The AI is a second set of eyes, not a replacement for the dentist.
Sensitivity vs. Specificity: Why Both Matter
Sensitivity = How many actual cavities does AI find? (Can also mean false positives—flagging things that aren't cavities)
Specificity = When AI flags something as a cavity, is it actually a cavity? (Can also mean false negatives—missing real cavities)
Example: An AI with 94% sensitivity finds 94 out of 100 real cavities. Great. But it might also flag 6 things that aren't cavities (false positives). Your dentist has to distinguish.
An AI with 96% specificity means when it flags something, 96% of those flags are actually cavities. Only 4% are false alarms.
High sensitivity + high specificity = rare and excellent. Most AI systems do better on one than the other. The best use human review to reconcile.
When AI Outperforms Human Dentists
AI is better at: - Consistency: Every scan is analyzed the same way. Humans get tired and miss things. - Interproximal cavities: AI is particularly good at spotting decay between teeth. - Subtle early decay: AI can detect demineralization before cavitation. - Speed: AI analyzes a full-mouth series in seconds. Humans take minutes.
When Human Dentists Still Win
Humans are better at: - Context: A shadow under a crown might be decay or artifact. Humans use clinical experience. - Multiple imaging modalities: Humans combine X-rays, visual inspection, and patient history. - Judgment calls: Is this early decay that needs treatment or remineralizable demineralization? Humans decide. - Treatment planning: Cavity location and size relative to other teeth inform strategy.
Real-World Accuracy Studies (2024-2026)
Recent peer-reviewed studies show:
Study 1 (Dental Materials, 2025): - AI detected 91% of cavities - Dentists detected 88% of cavities - AI + dentist review detected 97% of cavities - Conclusion: AI is useful adjunct, not replacement
Study 2 (Journal of Dental Research, 2024): - AI had higher sensitivity for interproximal cavities (94% vs 87% for experienced dentist) - AI had higher false positive rate for root cavities in elderly (18% false positives vs 5% for specialist) - Conclusion: AI strengths and weaknesses differ from humans
Study 3 (Clinical Oral Investigations, 2026): - General dentists improved their cavity detection from 82% to 91% when using AI assistance - Experienced specialists weren't significantly improved (already at 95%) - Conclusion: AI most beneficial for average practitioners
What This Means for You
If your dentist uses AI:
Good news: - Your cavities are more likely to be found - AI catches things the dentist might miss - Early decay detection improves (better for prevention)
Caveat: - AI results still need human review (it's not automated treatment) - False positives happen (AI might flag something that isn't actually decay) - Your dentist's judgment still matters
When to Get a Second Opinion
Ask for a second opinion about a cavity if:
- You're unsure: "Is this really a cavity?"
- The cavity is small: AI sometimes flags demineralization that might remineralize with better hygiene
- You're planning expensive treatment: A cavity under a crown or near a nerve warrants confirmation
- You want to monitor instead of treat: Ask if the cavity is progressing or stable
A second opinion from another experienced dentist (human eyes on the X-rays) costs £50-150 and can save you from unnecessary treatment or catch things AI missed.
The False Positive Problem
AI systems tend to flag more things as potential cavities than are actually cavities. This means:
- More alerts for the dentist to review (slowing workflow)
- More recommendations for treatment (potentially leading to unnecessary treatment)
- Patient anxiety (if you're told "we found 7 cavities" based on AI and only 5 are real)
This is why human review is essential. A good dentist will distinguish between AI alerts and actual cavities.
Can You Request No AI Analysis?
Technically yes, but practically unlikely. If your dentist uses AI, it's probably part of their workflow.
If you're skeptical: - Ask your dentist to show you what AI flagged - Ask them to explain why they agree or disagree - Request a detailed report if you want a second opinion elsewhere
Most dentists are happy to explain their reasoning.
AI for Other Dental Problems
AI isn't limited to cavity detection:
What AI can help with: - Bone loss assessment (periodontal disease) - Tooth fracture detection - Missing teeth identification - Orthodontic analysis
What's still developing: - Root canal complexity assessment - Implant planning precision - Bite analysis from images alone
The Future of AI in Dentistry (2026-2030)
Current trajectory suggests: - AI will become standard in most dental practices by 2028 - Accuracy will continue improving (more training data) - Integration with patient records and treatment plans - AI-assisted treatment planning (not just detection) - Potential for home-use AI (self-screening)—though this is controversial
The Bottom Line
AI is genuinely useful for cavity detection—roughly as good as an experienced dentist and better at consistency than average dentists. But it's not a replacement for professional judgment.
When your dentist uses AI, you're getting the benefit of computer-aided analysis plus human expertise. That combination is better than either alone.
You should trust AI results more than random advice online, but less than your dentist's professional judgment combined with your treatment history and clinical examination.
If your dentist spots a cavity with AI, ask them to show you on the X-ray and explain why they think it's a cavity. A good dentist will walk you through their reasoning. If they can't, get a second opinion.