Try OraCore Scribe free for 14 days. No credit card required.
Welcome to this week’s Dental AI Weekly, honest analysis of where dental AI is going, from someone building in it.
AI found signs of esophageal cancer hidden in routine CT scans. GE HealthCare and Mass General Brigham announced research into AI that could connect fragmented oncology records. Then hundreds of AI agents searched DNA in parallel and surfaced an enzyme system science had never characterized.
These are early signs of AI becoming a discovery engine.
Dentistry will not sit outside that change. Our radiographs, CBCTs, intraoral scans, clinical notes, treatment histories, saliva, and outcomes contain patterns no dentist could see across a single career. AI is beginning to give medicine a way to find them.
WHAT HAPPENED THIS WEEK
01: CANCER HIDING IN PLAIN SIGHT
An AI model found signs of esophageal cancer and precancerous lesions in routine CT scans that were taken for other reasons.
The September 22 Nature Medicine paper describes EAGLE, a model trained on 6,813 patients and tested across multiple hospital and screening settings. In external opportunistic-screening cohorts, it reached 98.5% specificity, 90.0% sensitivity for cancer, and 52.5% sensitivity for precancerous lesions. A prospective hospital validation processed 17,446 patients.
The striking part is where the information came from. These were routine noncontrast CT scans, not a new test created for the model. The signal was already there. AI learned how to see it.
Read the September 22 Nature Medicine study
What this could mean for dentistry: Dental AI already helps clinicians identify findings on radiographs. The larger possibility is earlier prediction. The same image may eventually reveal how disease is progressing, which patients are likely to deteriorate, or which combinations of oral and systemic findings deserve attention before the pattern is obvious to us. We may discover that routine dental imaging contains far more biological information than we knew how to read.
02: A PATIENT RECORD THAT CAN THINK ACROSS YEARS
GE HealthCare and Mass General Brigham are exploring generative AI that can bring a patient’s scattered radiation-oncology history together and make it available through questions.
The September 23 announcement describes research aimed at combining structured clinical data with the unstructured information buried across a patient’s treatment journey. GE also points to earlier workflow technology from Massachusetts General Hospital that reduced intake-to-treatment time from as long as 30 days to eight days across more than 11,000 treatment plans.
Read the September 23 GE HealthCare announcement
What this could mean for dentistry: Imagine opening a patient and having years of radiographs, periodontal measurements, treatment history, medications, missed treatment, symptoms, and clinical conversations already connected. AI could spot a change across time, explain why it matters, and prepare the likely next step before the dentist starts hunting through the chart. That could give the dentist a clearer view of how this patient’s condition has changed while removing the search and reconstruction that consume attention now.
03: AI FOUND SOMETHING WE HAD NOT FOUND
Anthropic says Claude agents searched DNA data and surfaced a previously uncharacterized enzyme system with CRISPR-like repeats.
About 950 AI agents used roughly 210 million tokens over 21 hours to investigate DNA sequences. They identified a candidate system built around a reverse transcriptase found in a jumbo phage. Laboratory experiments found that ART’s repeat array produces a set of distinct short RNAs. Anthropic calls the system array-associated reverse transcriptases, or ART.
The function of ART is still unknown. One possibility worth investigating is whether these RNA arrays could support a new programmable biological tool. The agents did not retrieve a known answer. They searched an enormous field of possibilities and found something science had not characterized before.
Read the September 23 Anthropic announcement
What this could mean for dentistry: Dental research has huge unanswered questions in oral biofilms, inflammation, salivary diagnostics, biomaterials, tissue regeneration, and the relationship between oral and systemic disease. AI agents could search those spaces continuously, combine findings across disciplines, and propose candidates for laboratory testing. The next important dental material, diagnostic marker, or regenerative treatment may begin with a pattern found by machines and proven by scientists.
“The most important thing AI may do for medicine is show us what we never thought to look for.”
THE BIGGER SIGNAL
The next generation could build a continuously improving model of oral health from imaging, clinical conversations, treatment response, medical history, and outcomes across millions of patients. It could help us understand why similar cases progress differently, identify disease earlier, design better materials, personalize prevention, and turn discoveries from medicine into dental care much faster than they move today.
Dentistry has spent decades producing valuable data and storing most of it in separate boxes. AI may finally let us learn from all of it.
AI OPTIMISM
21 hours
Claude agents examined more than 200,000 reverse transcriptases, identified 3,500 candidate systems, and narrowed them to 20 detailed reports.
That suggests a future in which researchers can investigate far more biological possibilities before choosing which experiments to take into the laboratory.
Source: Anthropic, September 23, 2026
BY THE NUMBERS
46%
of U.S. adults are not sure whether doctors or other healthcare providers have used AI in their healthcare.
What it signals: Nearly half of adults cannot say whether AI has been used in their care. Its growing role may be more visible to clinicians than to patients.
Source: Pew Research Center, August 25, 2026. Pew surveyed 3,488 U.S. adults from June 22 to 28, 2026.
READER Q&A
“How accurate are AI scribes during a real dental appointment?”
BH: Very accurate. The biggest thing I would tell any provider is that you do not need to change how you talk or start dictating every detail like a robot. Speak naturally and be clear when it matters. The clearer you are about the tooth, procedure, material, and next step, the more complete your note will be.
The system also lets you train your own templates, so what you say means what your practice intends. You can set it so “crown on 14” means D2740. If “cap number 14” means a porcelain crown, buildup, and your crown-seat workflow, your template can turn that shorthand into the correct codes and documentation.
OraCore also understands more than 100 languages and can still create the clinical note in English. The provider and patient can speak in the language that works best for them while the practice gets consistent English documentation.
The goal is not to make you work for the scribe. The system learns how you already speak and turns it into the note you need.
FROM THE ORACORE BLOG THIS WEEK
Evaluation | 8 min
How to Evaluate Dental Voice AI as It Moves Into Structured Charting
Structured charting is one early example of AI moving beyond transcription to organize spoken findings by tooth, surface, condition, treatment, and clinical context.
Try OraCore Scribe in the appointments you already run. Scribe Solo and Scribe Team include a free 14-day trial with no credit card required.
That’s the week. Reply with what you’re seeing in your own practice. I read every one.
Brad Hutchison
CEO, OraCore AI
Dental AI Weekly drops every Monday. pulse@oracorenews.com | OraCore AI, Denver CO