A landmark 2026 study published in the journal Radiology reveals that AI-generated "deepfake" dental and medical X-rays are now so realistic that experienced radiologists correctly identified them only 41% of the time — worse than a coin flip. For dental practices across Canada, this raises urgent questions about diagnostic integrity, insurance fraud prevention, and the security of digital imaging systems.
As of May 2026, artificial intelligence can generate synthetic radiographic images that are virtually indistinguishable from authentic clinical X-rays. A multi-centre research team, publishing their findings in the Radiological Society of North America's flagship journal, tested both seasoned radiologists and leading AI models against ChatGPT-generated X-rays — and the results should concern every dental professional who relies on digital imaging.
What the Study Found
Researchers asked radiologists with varying levels of experience to review a mix of genuine clinical X-rays and AI-generated synthetic images. Without being told that fakes were present, participants correctly flagged the synthetic images only 41% of the time. Even after being explicitly warned that some images were fabricated, accuracy climbed to just 75%.
Individual radiologist performance ranged widely — from 58% to 92% detection accuracy — but critically, years of clinical experience offered no advantage. A radiologist with 20 years of practice was no better at spotting fakes than a recent graduate.
The study also tested four leading multimodal AI models, including GPT-4o, GPT-5, Gemini 2.5 Pro, and Llama 4 Maverick. Their accuracy ranged from 57% to 85%, confirming that even advanced AI systems struggle to distinguish real from synthetic radiographs.
Why This Matters for Canadian Dental Practices
Digital radiography is the backbone of modern dental diagnostics. Intraoral periapical X-rays, bitewings, panoramic images, and cone-beam computed tomography (CBCT) scans guide treatment decisions every day in practices across Toronto, Mississauga, Brampton, Markham, and the broader GTA. If synthetic images can fool trained professionals, the implications extend far beyond academic curiosity.
Insurance Fraud and Claim Integrity
The Royal College of Dental Surgeons of Ontario (RCDSO) and the Canadian Dental Association (CDA) both require radiographic evidence to support certain treatment decisions and insurance claims. Fabricated X-rays could be used to justify unnecessary procedures, inflate claims, or manufacture evidence of conditions that do not exist. For insurance carriers processing claims under the Canadian Dental Care Plan (CDCP) and private benefit plans, this represents a significant verification challenge.
Pro Tip: If your practice receives digital referral images from unfamiliar sources, verify the DICOM metadata — authentic clinical images carry equipment-specific tags (manufacturer, model, acquisition parameters) that AI-generated images typically lack.
Patient Safety Risks
Deepfake radiographs could theoretically be injected into patient records through cyberattacks on practice management systems. Researchers have warned that hacking health systems to inject fake medical data could be more damaging than stealing data, because clinicians would not be able to distinguish which portions of the medical chart are genuine.
For dental practices in Ontario that must comply with the Personal Information Protection and Electronic Documents Act (PIPEDA), this adds another dimension to digital security obligations. Your imaging data is a clinical asset — not just a privacy concern, but a diagnostic integrity issue.
Legal and Forensic Implications
Dental radiographs play a critical role in forensic identification, malpractice defence, and regulatory proceedings. If the authenticity of digital images can be credibly challenged in court or before the RCDSO, it could undermine the evidentiary value of radiographic records across the profession.
How to Spot a Deepfake X-Ray
Researchers identified several recurring artefacts in AI-generated radiographic images that, while subtle, provide detection clues for trained observers:
- Unnatural symmetry: AI-generated images tend to produce overly symmetrical anatomical structures — real human anatomy is inherently asymmetric
- Overly smooth bone surfaces: Genuine bone trabeculation has a textured, irregular appearance that AI often smooths out
- Unnaturally straight spines: In panoramic and cephalometric views, AI tends to produce unrealistically aligned vertebral columns
- Uniform noise patterns: Clinical X-rays have sensor-specific noise profiles; AI images often show unrealistically even noise distribution
- Clean fractures and pathology: AI-generated pathological findings tend to appear too "textbook perfect" compared to the messy reality of clinical disease
Pro Tip: When reviewing referral radiographs, zoom to 200% and examine the bone-soft tissue boundaries. AI-generated images frequently produce unrealistic transition zones that become apparent at higher magnification.
What Your Practice Can Do Now
While this technology is still emerging, proactive steps can protect your practice and your patients:
1. Secure Your Imaging Chain
Ensure your practice management software (Dentrix, ABELDent, ClearDent, or similar) maintains audit logs for all imaging data. Any image added or modified should carry a traceable digital footprint. Discuss imaging security with your IT provider at your next review.
2. Verify Referral Images
When accepting digital referrals, confirm the source. If images arrive via email rather than through a secure referral platform, treat them with additional scrutiny. The Ontario Dental Association (ODA) has recommended secure digital communication channels for inter-practice referrals.
3. Maintain Original Acquisitions
Always retain original DICOM files from your imaging equipment. These carry embedded metadata that is extremely difficult to replicate in synthetic images. If a dispute ever arises about the authenticity of a radiograph, original DICOM data is your strongest evidence.
4. Stay Informed on Detection Tools
Several research groups are developing watermarking and provenance-tracking technologies specifically for medical imaging. The RSNA (Radiological Society of North America) is actively exploring standards for authenticating clinical images. Watch for Health Canada guidance on this issue as the technology matures.
Pro Tip: Add "imaging data integrity" as an agenda item at your next team meeting. Even a 15-minute discussion about verifying referral images and securing your DICOM storage can strengthen your practice's defences.
The Broader AI Integrity Challenge
This research arrives at a time when Canadian dental practices are rapidly adopting AI-assisted diagnostic tools — from caries detection algorithms to periodontal risk assessment models. The same generative AI capabilities that create convincing deepfake X-rays also power many of the diagnostic tools now entering clinical use.
The distinction matters. AI tools that analyse authentic clinical images to assist diagnosis are fundamentally different from AI that generates synthetic images. The former is a clinical asset; the latter, when misused, is a threat. Dental professionals should understand this distinction as they evaluate AI tools for their practice.
For dental practices in the Greater Toronto Area and across Ontario, the message is clear: digital imaging remains an indispensable diagnostic tool, but the era of assuming that a radiograph is authentic simply because it looks authentic is ending. Verification, security, and vigilance are the new requirements.
Frequently Asked Questions
Q: Can AI-generated deepfake X-rays be used for dental insurance fraud in Canada?
Yes, this is a growing concern. Fabricated radiographic images could theoretically be submitted to support fraudulent insurance claims under both private dental benefit plans and the Canadian Dental Care Plan (CDCP). Insurance carriers and regulatory bodies like the RCDSO are monitoring this emerging risk, and practices should maintain original DICOM files as proof of image authenticity.
Q: How can Canadian dentists protect their digital imaging systems from deepfake injection?
Dental practices should ensure their practice management software maintains comprehensive audit logs for all imaging data, use encrypted storage for DICOM files, require secure digital channels for referral images, and conduct regular cybersecurity reviews. Compliance with PIPEDA standards for digital health data provides a baseline, but imaging-specific security measures are becoming increasingly important.
Q: Are there tools available to detect deepfake dental X-rays?
As of May 2026, dedicated deepfake detection tools for dental radiographs are still in development. However, clinicians can look for telltale artefacts including unnatural symmetry, overly smooth bone surfaces, uniform noise patterns, and unrealistically clean pathological findings. The RSNA is actively developing authentication standards for medical imaging, and Health Canada is expected to issue guidance as the technology evolves.
EBIKO Dental will continue monitoring developments in AI imaging security and their implications for Canadian dental practices.