Artificial intelligence news.
Straight from the original announcements.

DentScribe Reports 100% Accuracy Across Eight Perio Charting Levels

November 10, 2026, 06:29Source: PR Newswire
DentScribe Reports 100% Accuracy Across Eight Perio Charting Levels
Image: PR Newswire

DentScribe has launched its AI Periodontal Charting Benchmark 1.0 and reported that DentScribe 2.165.0 passed all 68 tests across eight levels. The company said the system produced all 1,967 expected chart entries correctly, with zero extra entries.

Key points
  • DentScribe reported 100% accuracy across eight benchmark levels.
  • The system passed 68 of 68 tests and correctly produced 1,967 of 1,967 expected entries.
  • No extra chart entries were recorded in the main evaluation.
  • DentScribe also passed 13 noisy-room text versions with 471 of 471 entries correct.
  • The benchmark kit includes scripts, expected charts, and a score card for practices.

A benchmark built around periodontal charting

The benchmark was developed by DentScribe with teams of dentists, dental hygienists, and dental assistants. It combines a written charting standard, 68 test scripts, expected charts, and a score card for evaluating AI perio charting.

The standard covers universal tooth numbering, six sites per tooth, reading order as the probe travels around the mouth, and rules for missing teeth, corrections, findings, and scope. An entry is counted as correct only when the value or finding is assigned to the right tooth and site. A test passes only when every expected entry is correct and no extra entry appears.

Eight levels test navigation, findings and safety

The eight levels progress from individual readings to complete examinations. They test explicit tooth and site charting, numbers-only dictation, midline direction changes, arch transitions, and missing teeth. Other levels cover periodontal findings such as bleeding, suppuration, recession, mucogingival junction measurements, and mobility.

The benchmark also tests flexible phrasing, filler words, patient conversation, corrections, shorthand, scope words, tooth lists, and ranges. Its safety level checks that missing teeth receive no readings, incomplete readings remain incomplete, and findings do not spread to unmentioned sites. The final level requires two complete examinations, expressed in both full words and shorthand, to produce matching charts.

In its November 2026 results report, DentScribe said it handled all 24 tested dictation styles perfectly. The company also reported that all 81 clean and noisy scripts produced identical charts across eight runs per script.

Additional noisy scripts and a practice testing kit

Beyond the 68 main tests, DentScribe said it passed 13 additional noisy-room text versions, with 471 of 471 expected entries correct and zero extra entries. The scripts include simulated speech-recognition mishearings. DentScribe clarified that they are not a live suction-noise or microphone study.

For the second complete exam, shorthand reduced the script from 397 words to 210, about 47% fewer words. The company said this measures script length, not demonstrated chairside time savings.

DentScribe is making the kit and performance report available to practices. The kit includes read-aloud scripts, expected charts, and a score card. Practices can use the same reader, room, and microphone, score the first attempt, and count extra entries separately. The company said the benchmark is intended to let customers inspect how an AI system handles location, context, corrections, and changes in direction during periodontal charting.

DentScribeDentScribe 2.165.0AI Periodontal Charting Benchmark 1.0healthcarebenchmarksresearch findings

Related news

i-Lumen Reports Dry AMD Signals in i-SIGHT Feasibility Study
ResearchNovember 10, 2026, 15:17

i-Lumen Reports Dry AMD Signals in i-SIGHT Feasibility Study

i-Lumen Scientific reported findings from a randomized, sham-controlled feasibility study of its investigational AMD System, identifying potential improvements in visual…

TriNetX Adds Human-Validated Facts Extracted From Clinical Notes
Products and toolsNovember 10, 2026, 14:27

TriNetX Adds Human-Validated Facts Extracted From Clinical Notes

TriNetX has launched NLP-extracted Facts from Clinical Notes, a capability that gives researchers confidence-scored data points from clinical notes and links them to…

Inductive Wins OpenADMET CYP Challenge for Fourth Straight Time
ResearchNovember 6, 2026, 17:00

Inductive Wins OpenADMET CYP Challenge for Fourth Straight Time

Inductive placed first in the OpenADMET CYP Direct Inhibition Blind Challenge, finishing ahead of more than 270 participants from biopharma, academia and AI…