The Practice Ledger is a newsletter about the business of running a dental practice: the software, the money, the contracts and the staffing. It is written for owner-dentists, for associates who are about to become owners, and for the office managers who have to make any of it work on a Tuesday morning. We cover the operational side of dentistry, not the clinical side. The question we ask about everything is narrow and unglamorous, and it is this: is the thing worth your money and your chair time?

I am Thad Connelly, and I write every word of it. There is no staff, no sponsor and no sales team behind this. For the guide below, that meant reading what the vendors claim, checking those claims against the FDA clearance database, and then going to find the dentists and hygienists who have already paid for the software and asking whether it survived contact with a real schedule.

What we are not: a review site with an affiliate feed, a consultancy that will later sell you the implementation, or a trade title that runs vendor copy as editorial. We are nobody’s channel partner and nobody’s reseller. We are also not your lawyer, your accountant or your clinical adviser. The disclosure below is not boilerplate, and it is worth the two minutes.

This issue is a field guide to every dental AI product we could find as of August 2026, sorted by the job it actually does: what it is, what it costs where anyone will admit to a price, and what the people already using it say when there is no sales rep in the room.

Four categories, in the order most practices meet them. Where a company sells into more than one, it shows up more than once.

Pearl runs the broadest suite: Second Opinion for 2D radiographs, Second Opinion 3D for cone beam, plus Practice Intelligence, Claimcheck, Calibrate and the new Pearl Voice. It claims 94% detection accuracy and a read in under 30 seconds. The ADA made a strategic investment in Pearl in 2024.

Overjet holds the deepest stack of FDA clearances in the category, covering caries, periapical radiolucency, bone level, pediatric use and image enhancement. It also sells its own imaging software, IRIS, insurance verification across more than 300 payers, and a new Voice AI suite.

VideaHealth ships Videa Caries Assist, Videa Dental Assist and interproximal bone-level measurement, and is embedded across Heartland Dental and Henry Schein One. It raised $40 million in 2025.

Denti.AI sells Detect and Auto-Chart. Auto-Chart was the first dental auto-charting product the FDA cleared.

Velmeni added FDA clearance for 3D cone beam analysis in July 2026, the newest clearance in the category.

Also selling into imaging today: Dexis DTX Studio Clinic, Diagnocat, Cube Click and Kells.

Bola AI is the most widely deployed voice tool in dentistry: Voice Perio, Voice Restorative and an AI Scribe, integrated with the major practice management systems through Henry Schein and Curve. It claims 99% accuracy across more than seven million exams.

Denti.AI, the company behind Detect and Auto-Chart, also sells a voice perio product and a scribe that handles more than 20 languages and writes notes straight back into the practice management system.

Newer and much less proven: Alta AI, DentScribe, Overjet's Voice AI Suite and Pearl Voice. Dentrix Voice and FP VoiceWorks are voice actuation rather than AI, which is not a criticism. For perio numbers it may be all you need.

Arini is the dental-specific AI receptionist with the loudest case studies. It takes inbound and after-hours calls, runs outbound reactivation campaigns and books straight into the practice management system. A single-site practice can be live in days.

Weave bought TrueLark in May 2025 for $35 million and bolted agentic AI onto its phone and messaging platform. Plans start at $199 a month and include onboarding and one-to-one training, which matters a great deal for our 30-minute bar.

Dental Intelligence added an AI Receptionist and an ask-your-data assistant on top of its analytics platform. It says the average practice added $180,000 in the first year, across a sample of roughly 3,000 practices.

Also in the category: Peerlogic for call analytics, Patient Prism for call coaching, plus Viva AI, Rondah AI and HeyGent.

Overjet verifies coverage days before the appointment, with code-level breakdowns across more than 300 payers.

Pearl Claimcheck scrubs claims before they are submitted, and Vyne Trellis handles claims, attachments, eligibility and remittances in one cloud platform.

Zuub is genuinely strong at verification, with more than 350 payers and a claimed 95% accuracy, but it has repositioned itself as infrastructure with a developer API aimed at large groups and software companies. That is an integration project, not a 30-minute training, so it fails our day-one bar for a small practice.

Zentist and Dental Intelligence also sell into this lane.

Start with an uncomfortable fact. There are almost no independent reviews of clinical dental AI. Overjet's G2 profile carries a single review. Pearl has four on Capterra. Weave, which is fundamentally a phone system, has close to 700 on Capterra and more than 400 on G2, and Dental Intelligence has roughly 50 on each. The clinical products have effectively no verified review base at all, and a large share of the "best dental AI" articles ranking on Google are written by companies selling in the same category. One review site we tried to read had already been taken offline.

So the real reviews live on Reddit and in Facebook office-manager groups. We read them. Four things come up again and again.

False positives are the number one complaint, on every imaging vendor. A practice evaluating all three of the big platforms this summer described them as riddled with false positives and negatives, and suspected the tools were built to help large groups justify treatment. A Pearl user in the same thread called the diagnostics merely okay. A VideaHealth user said they now clear false findings off the image before walking into the room, and thought accuracy had improved over six months.

The reliable return is case acceptance, not diagnosis. Almost every positive comment we found was about the patient understanding their own X-ray. Several dentists called it a useful tie-breaker on borderline caries and a real help on periapical radiolucency, then added that they would think hard before paying for it out of their own pocket.

Voice charting works, for the narrow job only. The best review we read came from a hygienist charting perio by voice into Eaglesoft. Numbers, bleeding, suppuration and recession all chart reliably, mistakes are easy to undo, and the commands take minutes to learn. Two warnings from the same review. If the patient says a keyword mid-exam the software will chart it, so tell them the robot is listening. And the auto-generated patient treatment summary was, in their view, unreliable enough that they close it before the patient can see it.

Your operatory matters as much as the software. Microphone quality and room noise came up repeatedly, along with a genuine learning curve measured in days, not minutes. Imaging AI degrades on imperfect films too. One dentist specifically flagged unreliable reads at the cervical margin because of burnout.

Two things a sponsored guide would not tell you.

The insurers are on the other side of the table. Overjet is backed by Delta Dental and acquired American Dental Examiners, whose review work sits behind claim decisions at several major carriers. The engine that helps you diagnose can help a payer deny. Dentists raise this unprompted, and it is a fair question to put to a sales rep. The ADA's own investment in Pearl raises a milder version of the same question.

There is a live overtreatment story. Former Wall Street Journal columnist Joanna Stern reported in her 2026 book I Am Not a Robot that a practice using dental AI recommended a multi-visit periodontal plan costing thousands of dollars. Several other dentists disagreed with the finding and she never had the treatment. The published reporting names Pearl as the software in use at that office. Employees at dental offices told her their bosses leaned on them to turn AI findings into booked work. Whatever you make of it, your patients are reading it.

Our own position, for what it is worth. Buy these tools to communicate better and to stop leaking money on the phones and on claims. Do not buy them to diagnose for you, never let a colored overlay create a treatment plan on its own, and never let anyone in your office be measured on how much the software sold.

That is every product we could find this month, and the list will not hold for long. The clearances are the part most likely to change first.

— Thad

Almost every vendor in this guide answers the price question with "contact us." The next issue works that problem: what practices are actually paying per operatory and per doctor, which contracts auto-renew and on how much notice, who bills per scan or per chart, and what it costs to walk away. Same rules as this one — no sponsors, no affiliate links, and no number printed as fact unless a reader or a published source stands behind it.

If you are paying for anything named above, reply and tell me what you pay. I will publish the ranges and never the practice.

The Practice Ledger lands in your inbox every week. One email, free, no sponsors, and one click to leave if it ever stops earning the time. If this was forwarded to you and you want the next issue, sign up below.