Pearl's FDA-cleared dental AI analyzed 26 million dental visits producing 737 million individual tooth observations across 15 million patients in all 50 U.S. states and the District of Columbia. The picture it reveals is dramatically different from what public data has told us for decades. Every patient in this dataset has already walked through a dentist's door — these are best-case numbers for the mouths the health system is actually reaching.
America's cavity problem is at least 4.5× worse than federal NHANES data says. The CDC's NHANES relies exclusively on visual-tactile exam (mirror, explorer, compressed air) to estimate that roughly 21% of American adults have untreated decay. Pearl's AI reads the radiographs NHANES deliberately excludes and finds 8.7× more decay per patient — Pearl 6.07 decayed teeth per patient vs. NHANES 0.7. The two methods are complementary; NHANES uses only one. Missing teeth, which both methods can equally observe, nearly match (Pearl 2.16 vs. NHANES 2.0), as do restored teeth (Pearl 7.01 vs. NHANES 6.0) — corroboration that lends credence to the much larger gap in decay detection.
In the most underserved zip codes, patients lose 40% more teeth. Pearl matched zip-code data against the National Provider Identifier registry of 180,974 U.S. dental practices and 295,509 dentists. In the 25 U.S. zip codes with no resident dentists (5,715 patients), 20.8% of affected teeth are extracted versus 14.9% in zip codes with 51+ dentists. The effect is concentrated at the extreme — the middle density tiers cluster between 13–16% — but where access disappears, treatment shifts decisively toward extraction rather than restoration.
Same zip code, same community, radically different diagnoses. Across 937 zip codes where Pearl could compare two or more offices (each with 100+ adult patients), the typical within-zip spread in untreated-decay rate is 9.2 percentage points; the top 10% of zips show a P90 spread of 17.8 pp; the largest within-zip spread observed is 45 pp. Across 282 dental offices each seeing 200+ adult patients, the 90th-percentile practice finds 4× more affected teeth per patient than the 10th-percentile practice (12.4 vs. 3.1). Pearl's AI analyzed every x-ray using the same FDA-cleared model — what varied was how offices responded to what it found.
Young adults have the most neglected mouths; older adults have the most managed. Total disease rises from 13.2 teeth affected at 18–24 to 19.1 at 75+. But the share left untreated falls from 52% (18–24) to 28% (75+). Restorations overtake untreated decay at the 35–44 bracket — the lifetime crossover.
Four teeth carry a third of all restorations. First molars (#3, #14, #19, #30) are 14% of the mouth but account for 24% of all decay and 31% of all restorations. Median age of first decay detection by tooth group: First Molars 17, Second Molars 18, Premolars 18, Canines 21, Upper Incisors 21, Lower Incisors 23 — a six-year prevention window from the most vulnerable position to the most resistant.
Hawaii has the lowest disease burden (9.63 affected teeth). Delaware has the highest (14.22). New Jersey leads in calculus prevalence (67.7%); Kansas leads in extractions. National averages: 15.24 affected teeth per patient, 39.8% untreated rate, 45.8% restored rate, 72.9% calculus prevalence.
Most of this report is about the United States, but Pearl's data covers the U.K. NHS too. On two questions about American oral health — who makes it into the chair, and what happens when disease is found — the British system provides a clarifying contrast, used as a comparative lens rather than a parallel story.
American women are 28% more likely to visit the dentist than men (56.2% female vs. 43.8% male in the U.S. dataset). Under the UK's NHS, the split is exactly 50/50. Across 9,517,165 U.S. patients with known sex, clinical numbers are remarkably similar between men and women once they're in the chair (decayed teeth: 3.87 vs. 4.03; restored: 7.97 vs. 7.37; untreated rate: 31.4% vs. 33.9%). The gap isn't in the mouth — it's in who makes it through the door.
When disease is found, Americans restore; the British extract. U.S. fee-for-service insurance incentivizes restoration; NHS budget constraints make extraction more accessible. U.S. FMX cohort: 6.07 decayed, 2.16 missing, 7.01 restored teeth. U.K. FMX cohort: 7.79 decayed, 6.36 missing, 6.41 restored — a 3× gap in missing teeth. 14.5% of affected teeth are extracted in the U.S. versus roughly 31% in the U.K. FMX cohort.
Every patient here already made it through the access barrier. The true U.S. oral-health burden — including those who can't or don't get care — is almost certainly worse than what Pearl's numbers show. Across 2,033 zip codes matched to Census income data, we find essentially no correlation (−0.06) between income and dental outcomes within the care-engaged population. Within the same population, the share of untreated decay in fluoridated communities (39.6%) is statistically identical to non-fluoridated (37.7%). 36% of decay detections aren't reproduced at the next visit — some real clinical change, mostly AI inconsistency at detection margins, which we report honestly.
The Pearl Oral Health Index draws on two complementary datasets covering adults 18+, all 50 U.S. states and the District of Columbia, plus the United Kingdom, using a 28-tooth universe (excluding wisdom teeth), across April 2024 through March 2026. Dataset 1, FMX Series: 1.49M patients analyzed with full-mouth x-ray series — the authoritative source for per-patient averages. Dataset 2, Expanded Encounters: 14.8M patients with 26 million tooth-level findings, powering state-level analyses, gender comparisons, tooth-by-tooth vulnerability timelines, and 4.26 million longitudinal tooth transitions.
Pearl is a dental AI company. Founded in 2019 by a team with decades of experience in enterprise computer vision, Pearl introduced the first-ever FDA-cleared AI capable of reading and instantly identifying diseases in dental x-rays. With regulatory clearance in 120 countries and 50,000+ clinicians using its platform, Pearl's products include Second Opinion, Second Opinion 3D, Practice Intelligence, Precheck, Calibrate, and Pearl Voice. Request a demo.
Media contact: Nick Garrison — pr@hellopearl.com