AI Dental X-Ray Triage 2026: Pearl vs Overjet Payouts

Dental AI stopped being a demo the moment insurers started running it. FDA-cleared dental AI radiograph software from Pearl and Overjet now sits on both sides of the transaction — reading the bitewing on your operatory monitor and re-reading that same bitewing inside the payer’s claim adjudication queue. For practice owners, that changes the math from “cool visualization that boosts case acceptance” to “the algorithm that determines whether my crown claim gets paid in eleven seconds or sits in manual review for eleven days.” If you own a practice in 2026 and haven’t audited which AI your top three payers use, you are negotiating blind against a system you can actually study.

What’s actually new about dental AI radiograph software

The technical clearances are not the news anymore. Pearl’s Second Opinion received FDA 510(k) clearance for detecting multiple pathologies on adult radiographs, and Overjet holds clearances covering caries detection and bone level measurement. Both are legitimate FDA cleared caries detection AI products with published sensitivity figures that generally beat an average clinician reading quickly between patients. That part has been settled for a couple of years.

What changed is the distribution channel. Overjet built its business on the payer side first. Delta Dental plans across multiple states, Guardian, and other carriers deployed Overjet to automate claim review, particularly for periodontal scaling and root planing (D4341/D4342), where AI bone level measurement dental analysis can confirm or reject the bone loss that justifies the procedure. Pearl moved the other direction, saturating the practice and DSO market with chairside detection before pushing into payer integrations and third-party review. The result in 2026 is a small number of models with enormous leverage: your submitted radiograph gets scored by software whose thresholds you did not set and cannot see.

The information asymmetry runs both ways. Payers using Overjet dental AI can auto-approve clean cases in seconds — a real win when your documentation is strong — and auto-flag the marginal ones you used to win on narrative alone. Practices running Pearl Second Opinion review chairside can pre-screen their own submissions against comparable measurements before the claim goes out. Whoever runs the better pre-submission QC wins the reimbursement argument. That is a workflow problem, not a clinical one, and it is solvable this quarter.

Why it matters

• Your denial rate is now partly a data-quality problem. Underexposed, cone-cut, or improperly angled radiographs degrade AI confidence scores. Images a human reviewer would have accepted with a narrative can now fail an automated threshold before a human ever looks.
• SRP and crown claims carry the most exposure. Perio claims hinge on measurable bone loss from CEJ to crestal bone — exactly what AI bone level measurement dental tools quantify in millimeters. Subjective clinical justification loses to a number.
• Case acceptance goes up, but for a boring reason. Practices report meaningful lifts in treatment acceptance when patients see annotated overlays on their own films. Dental practice case acceptance AI works because it converts your professional opinion into something the patient can literally see, not because the model is smarter than you.
• Associate and hygiene calibration becomes auditable. When every operatory runs the same detection model, variance between providers stops being anecdotal. That is useful for coaching and uncomfortable for anyone who has been over- or under-diagnosing.
• DSO valuations are starting to reference it. Buyers increasingly ask whether diagnostic workflows are standardized and whether claim denial rates are documented. AI-backed consistency is becoming a diligence line item.
• The malpractice question is unresolved. If the model flags a lesion and the chart shows no follow-up, that discrepancy is discoverable. Running dental insurance claim AI tooling creates a record — decide your documentation policy deliberately, not by accident.

How to use dental AI radiograph software today

1. Pull your denial baseline before you buy anything. Export the last 12 months of claims from your PMS and segment denials by CDT code. Without this number, you cannot prove ROI on any AI purchase. In Open Dental (MySQL backend), a read-only query gets you there:

   SELECT pl.CodeNum, pc.ProcCode, pc.Descript,
          COUNT(*) AS total_claims,
          SUM(CASE WHEN cp.Status = 4 THEN 1 ELSE 0 END) AS denied,
          ROUND(100.0 * SUM(CASE WHEN cp.Status = 4 THEN 1 ELSE 0 END) / COUNT(*), 1) AS denial_pct
   FROM claimproc cp
   JOIN procedurelog pl ON pl.ProcNum = cp.ProcNum
   JOIN procedurecode pc ON pc.CodeNum = pl.CodeNum
   WHERE cp.DateCP >= DATE_SUB(CURDATE(), INTERVAL 12 MONTH)
   GROUP BY pl.CodeNum, pc.ProcCode, pc.Descript
   HAVING total_claims >= 10
   ORDER BY denial_pct DESC;

2. Find out which AI your payers actually run. Do not guess. Send this to your provider relations rep at each of your top payers by volume:

   Subject: AI-assisted claim review - provider inquiry, TIN [your TIN]
   
   Hello,
   
   We are updating our radiographic documentation protocols. Please confirm:
   
   1. Does your plan use automated or AI-assisted image analysis in
      adjudicating claims for D4341/D4342, D2740, or D3310?
   2. Which vendor or platform performs that analysis?
   3. What radiographic criteria (bone loss thresholds in mm, image
      type, date recency) trigger automated approval vs. manual review?
   4. Is there a provider-facing appeal path specific to
      AI-flagged determinations?
   
   Please respond in writing. Thank you.

3. Run a blinded pilot, not a vendor demo. Pick 100 recent bitewings and FMX series where you already know the treatment outcome. Have the vendor process them and compare against your charted findings. Track agreement, false positives on restoration margins, and — critically — how the tool handles your specific sensor brand and exposure settings.
4. Standardize image capture before deployment. AI confidence scores fall off a cliff with inconsistent geometry. Post this at every operatory and make it part of hygiene onboarding:

   RADIOGRAPH QC STANDARD - all operatories
   - Rinn XCP holders on every BW and PA. No fingers, no exceptions.
   - Bitewings: full interproximal contact open, no overlap on 1st/2nd molar
   - Verticals for any perio patient with 5mm+ probing
   - Exposure: [your sensor's kVp/mA/sec] - do not adjust per-patient
   - Reject and retake: cone cut, motion blur, contact overlap >1mm
   - FMX recency for SRP submission: within 6 months of service date

5. Pre-screen claims against the AI before submission. This is the whole game. If your chairside tool reports bone loss below your payer’s stated threshold, either add supporting documentation (perio charting, narrative, prior films showing progression) or reconsider the code. Build the check into your claims workflow as a hard gate:

   PRE-SUBMISSION GATE - D4341 / D4342
   [ ] Radiographs dated within 6 months, vertical BWs or FMX
   [ ] AI bone level measurement attached, >= payer threshold
   [ ] Full-mouth perio chart, 6-point probing, same date
   [ ] Radiographic calculus visible OR documented in narrative
   [ ] Quadrant tooth count meets code requirement (4+ teeth = D4341)
   [ ] If AI measurement is borderline: attach prior films + progression note
   Gate fails -> route to doctor review before submission, not after denial.

6. Write the discrepancy policy down. Decide in advance what happens when the AI flags something the doctor disagrees with. The defensible answer is a charted note stating the finding was reviewed and clinically ruled out, with a reason. Silence is the worst option.
7. Track the metric that pays. Ninety days post-deployment, re-run the query from step one. Look for movement in denial percentage on your top three codes and in per-patient treatment plan value. If neither moved, you bought a screensaver.

How Pearl and Overjet compare

Factor	Pearl (Second Opinion)	Overjet
Primary market entry	Practices and DSOs first; expanding into payer and third-party review	Payers and insurers first; strong practice-side push since
FDA posture	510(k) cleared, multi-pathology detection on adult radiographs	510(k) cleared, caries detection and bone level measurement
Signature strength	Breadth of detected conditions and chairside patient-facing overlays	Quantified bone level measurement in millimeters; perio claim workflows
Payer relevance	Growing; used in claims review contexts	Deep — deployed inside major carrier adjudication pipelines
Best fit	Practices prioritizing case acceptance and diagnostic consistency	Practices with heavy perio volume and payer-driven denial pain
Integration	Broad imaging and PMS integrations, cloud and on-prem options	Broad imaging integrations; analytics layer over practice performance
Pricing model	Per-location subscription, typically tiered by chair or volume	Per-location subscription; enterprise pricing for groups

Neither is obviously better. The honest selection criterion is which one your payers use and where your denials cluster. Heavy perio practice fighting Delta on SRP? Match your pre-screen to the tool doing the reviewing. Fee-for-service practice where case acceptance drives the P&L? Optimize for the patient-facing experience instead. Get both on a blinded pilot with your own films and let the numbers decide.

What’s next

Expect the measurement layer to become the contract layer. Once a carrier can state a numeric bone loss threshold and verify it automatically, that threshold migrates into the provider agreement itself. Practices will negotiate against published criteria rather than adjuster discretion — arguably an improvement, since a documented number is appealable in a way that “clinical judgment of the reviewer” never was. The practices that win will read the criteria and build capture protocols around them.

Watch for scope expansion beyond 2D radiographs. CBCT analysis, 3D segmentation for implant and endo planning, and intraoral scan integration are all in active development across the category. As those clear regulatory review, the same payer dynamic follows: a model that measures bone volume for an implant site will eventually be asked whether that site justified a graft. Assume every new imaging modality eventually gets a reimbursement gatekeeper attached.

Third is consolidation and platform lock-in. Both vendors are layering practice analytics, referral tracking, and revenue cycle tools on top of detection. That is a rational business move and a real risk for owners: the more your claim workflow depends on one vendor’s scoring, the weaker your position at renewal. Negotiate multi-year terms with data portability language now, before switching costs get expensive. Keep raw DICOM exports under your control.

Frequently Asked Questions

Does dental AI radiograph software actually reduce claim denials?

Indirectly, yes — but not because the AI argues on your behalf. It reduces denials when you use it to catch weak documentation before submission. Practices that deploy detection chairside and change nothing about their claims workflow typically see case acceptance gains without denial improvement. The denial win comes from the pre-submission gate, not the software license.

Is it a conflict of interest that insurers use the same AI I bought?

It is a legitimate tension, and worth raising with your state dental association. Practically, it also cuts in your favor: if you know the model and the threshold, you can document to it. That is far more actionable than the old system where denial criteria were effectively unknowable. Treat shared tooling as intelligence, not just as a threat.

How much does Pearl or Overjet cost per location?

Both sell per-location subscriptions and neither publishes rates, so pricing depends on group size, chair count, and contract length. Expect a monthly per-location fee in the range of a mid-tier practice software subscription, with meaningful discounts at DSO scale. Get quotes from both simultaneously and mention it — competitive pressure in this category is real.

Do I have to tell patients that AI reviewed their radiographs?

Disclosure requirements vary by state and no single national rule governs it, so confirm with your state board and malpractice carrier. Most practices disclose voluntarily because the annotated overlay is the point — patients respond well to seeing the finding highlighted on their own film. Frame it as a second set of eyes supporting the doctor’s diagnosis, which is accurate.

What happens when the AI and my clinical judgment disagree?

Your judgment governs — these are clinical decision support tools, not diagnostic authorities, and the FDA clearances are written that way. What you must not do is ignore the discrepancy silently. Chart that the flagged finding was reviewed and the clinical reason it was not treated. That note protects you far more than the absence of a note ever will.

Will FDA cleared caries detection AI replace the dentist reading the film?

No, and that framing misses where the value is. These models are strong at consistent detection of specific findings on specific image types and weak at everything requiring context — patient history, caries risk, restorability, what the patient will actually accept and pay for. The realistic outcome is that reading films gets faster and more consistent while the judgment layer stays entirely human.