Dentists want AI for charting, not diagnosis

Dentists are adopting AI in exactly one direction: paperwork first, judgment last. The ADA's own survey data shows the profession handing AI its administrative burden while refusing — by huge margins — to let it near treatment decisions.

I was just recently reading the Health Policy Institute's perspective. They conducted a survey on how dentists use AI, and the majority of AI use is still in the initial stages of implementation. This is in the US, where frontier models are available and a lot of fast-growing implementation is taking place.

The numbers

The results show that imaging and diagnostics currently account for only 22.8%, and insurance verification for 13.6%. Business intelligence stands at 10%, and explaining clinical findings to patients is at 13.2% — which also has a lot of potential. Social media use is around 10%. This data is from Q2 2026, so keep in mind this is still early.

The newer wave of the same ADA HPI survey shows the direction of travel: 25.4% of dentists now use AI for imaging and diagnostics, 30.6% plan to start using it for charting and notes, and 24.4% plan to adopt it for insurance tasks by the end of 2026. Adoption is climbing — but only on the administrative side.

Where the real potential lies

The most interesting part I was looking at is charting and note-taking — that's where I think the real potential lies, but dentists have yet to adopt it much.

The market agrees. Dental AI scribes have become a real category: 2026 buyer's guides now compare nine-plus tools, practices report saving 1–2 hours a day on documentation, and the most-adopted platform claims over 10,000 dentists and 3 million completed charts. Thirteen companies now hold FDA clearances for dental AI products. When a tool gives a clinician an hour of their day back, adoption takes care of itself.

The money is following the same trail. Investors are now backing dental AI companies that do nothing but insurance claims automation — nobody raises capital to replace a dentist's judgment, they raise it to clear the paperwork faster than the practice next door.

What dentists won't hand over

The majority of dentists — around 82% — don't want to use or don't plan to use AI for treatment recommendations. Similarly, 71% don't want to use it for appointment scheduling. In the current survey wave, fewer than 5% actually use AI for treatment recommendations.

That's not a technology gap. That's a trust boundary, and clinicians keep restating it plainly:

AI can analyze the scan, but the doctor makes the decision.

What patients think

There's a third party in this conversation the survey didn't poll: patients. They're already forming opinions, and not gently. When AI-generated dental "surgery" videos started flooding social feeds this year, a single dentist's reaction video picking them apart pulled over 4.3 million views — and the comment section was merciless about anything that looked machine-made.

That's worth taking seriously. Patients don't distinguish between a practice that uses AI to write chart notes and one that uses it to make their diagnosis. If they see "AI" and "dentist" together in the wrong context, trust drops before you get a word in. The lesson isn't to hide the tools — it's to be precise about what they do: the software handles the record, your dentist handles you.

Here's the catch

I'm not sure whether the slow charting adoption is because the technology is new, because dentists are skeptical about it, or because it's hard to implement within a setup that already has an established process. Probably all three.

And the skepticism has legs. Studies report tooth-level detection accuracy of 94.9–99.9% on radiographs, while some dentists report the same tools feel closer to 30% useful in the chair. Both can be true — benchmark accuracy and chairside usefulness are different things. The clinical literature is also now naming a real risk: AI sensitive enough to flag findings too mild to ever need treatment, which is overdiagnosis wearing a lab coat.

What to do with this

If you run a practice, don't start with diagnosis — start where the survey says your colleagues are quietly winning. Pick one administrative workflow (charting is the obvious candidate) and trial an AI scribe on your own notes for two weeks. Measure minutes saved, not accuracy claims.

Two numbers tell you everything after those two weeks: minutes of documentation per patient, and how many notes are finished before you leave the practice. If both move, keep the tool — and only then look at the next workflow. If they don't, you lost fourteen days and a trial fee, which is exactly the kind of failed experiment a practice can afford.

Whatever you adopt, say it plainly on your website and in the chair. Patients accept software that types while the doctor thinks — but only when it's explained before they ask.

So it's an interesting picture: dentists show the most interest in adopting AI technology to assist with charting in the dental sector. There are other publications on this topic that I'll share later — this is just the beginning article.