Surface Story
An investigation has found that dental practices are increasingly using AI diagnostic tools to identify cavities, gum disease, and other conditions in patient X-rays. Critics, including former dental professionals and patient advocates, allege that some practices are using AI outputs to recommend and perform unnecessary procedures, generating revenue from treatments patients do not need. The pattern appears to be concentrated in corporate dental chains.
Underneath the Story
The dental industry has always operated on a fundamental information asymmetry: patients cannot see inside their own mouths, cannot read their own X-rays, and have no independent way to verify a diagnosis. That asymmetry has always created the conditions for overtreatment, and overtreatment has always existed in dentistry. AI does not create this problem. It gives practices a new tool to make overtreatment feel objective, algorithmic, and therefore unquestionable. An AI flagging a shadow on an X-ray as a cavity has the authority of a machine. Arguing with it feels like arguing with a fact.
AI didn't give dentists the ability to recommend unnecessary procedures. It gave them a way to make those recommendations feel like they came from a computer instead of a person with a financial interest.
The information asymmetry in dental care is structurally total. You sit in a chair with your mouth open. A trained professional examines surfaces you cannot see, reads images you cannot interpret, and renders a diagnosis you have no independent means of verifying. Then they tell you what it will cost to fix it.
This has always been the design of the dental encounter. It is not unique to dentistry: medicine, auto repair, and legal services share the same basic structure. An expert with specialized knowledge diagnoses a problem the customer cannot evaluate and recommends a solution the customer cannot independently assess. The customer's options are trust, seek a second opinion, or decline treatment. Most people trust.
What AI adds to this arrangement is a specific kind of authority that is new and that the existing dynamics of the dental encounter are not equipped to handle.
The Second Opinion Problem
The structural protection against overtreatment in any expert-knowledge industry is the second opinion. If a dentist recommends a crown you are not sure you need, you can go to another dentist and ask them to look at the same X-ray. Second opinions in dentistry work imperfectly but they work. Dentists disagree. Studies have found significant variation in treatment recommendations for identical X-rays across different practitioners. That variation is itself evidence that diagnosis in dentistry involves judgment, not just objective measurement.
AI outputs are positioned, by the practices that use them and by the general cultural authority currently attached to artificial intelligence, as something categorically different from a practitioner's judgment. An algorithm flagging a lesion on an X-ray feels less like an opinion and more like a reading. It has the aesthetic of objectivity. The interface typically looks like a scan result, with highlighted regions and confidence scores, not like a recommendation from a person with a financial incentive.
When a patient hears "our AI system identified a cavity forming here," the implicit message is that a machine has seen something. Machines, in the popular understanding, do not have financial incentives. They do not get paid per procedure. They just process the data and report what they find.
This is not accurate. The AI was trained on data chosen by someone. It was optimized toward an output defined by someone. The threshold at which it flags a shadow as a cavity worth treating was set by someone. All of those decisions were made by people with interests, and in the dental AI industry, those interests include selling software to practices that profit from higher treatment volumes.
The Corporate Chain Context
The investigation focuses on corporate dental chains rather than independent practices. This is structurally significant. Independent dental practices have reputational incentives that constrain overtreatment: a dentist who consistently recommends unnecessary work loses patients through word of mouth in the community where they practice. The relationship between dentist and patient has historically included accountability through ongoing personal relationship.
Corporate dental chains operate differently. They have high patient turnover, centralized management with revenue targets, and compensation structures that reward production volume. The individual dentist practicing within a corporate chain faces institutional pressure that the independent practitioner does not. AI tools that increase flagging rates and treatment recommendations serve the corporate chain's financial model regardless of whether they serve individual patients.
The AI is being deployed into an institutional environment that was already incentivized toward overtreatment. It does not create the incentive. It provides a tool that makes acting on the incentive feel more defensible and look more legitimate.
What Objective Authority Does to Informed Consent
Informed consent is the ethical and legal foundation of medical treatment. A patient must understand what is being recommended, why, what the alternatives are, and what happens if they decline. For informed consent to function, the patient must feel genuinely free to ask questions, seek other opinions, and decline treatment without social pressure.
The algorithmic authority of AI diagnosis subtly undermines this. When a machine has "identified" a problem, declining treatment feels less like an informed choice and more like ignoring a finding. The patient's subjective experience of the diagnosis shifts from "this dentist thinks I have a cavity" to "there is a cavity." The first framing invites questions. The second forecloses them.
Practices that present AI diagnoses this way, as findings rather than assessments, are changing the informed consent dynamic without changing the words they use. The patient is told what the AI found. They are not told how the AI was trained, what its false positive rate is, who built it, or what financial relationship exists between the practice and the AI vendor.
The Verification Gap
The underlying problem is that most patients will never be able to verify whether the procedure they received was necessary. If you get a crown placed on a tooth the AI flagged, you will not subsequently learn whether that tooth actually had a cavity forming. You got treatment and now you have a crown. The counterfactual, what would have happened if you had waited or sought a second opinion, is permanently unavailable.
This verification gap has always existed in dentistry. It is why the information asymmetry is so durable and why overtreatment is so difficult to detect at the individual level. The patient who received an unnecessary crown does not know it was unnecessary. They know they went to the dentist, something was found, they got it treated, and now they are fine.
AI makes this gap wider by adding a layer of apparent objectivity to the diagnosis that makes second opinions feel less necessary and makes patients less likely to seek them. The tool that should increase diagnostic accuracy is being used, in at least some institutional contexts, to increase diagnostic authority in a way that serves the institution more than the patient.
The dentist has always known things you cannot verify. Now they have a machine that agrees with them.
Editorial Note
underneath.news analyzes structural patterns, power dynamics, and the conditions that shape contemporary events. This is original analytical commentary, not reporting. We do not summarize, paraphrase, or replace coverage from any specific publication.
