AI or Not, a prominent leader in AI-generated content detection, has released results from an independent benchmark using a curated dataset from the March 2026 Radiology study, "The Rise of Deepfake Medical Imaging." In a blinded test, AI or Not's detection technology achieved 95% overall accuracy and correctly identified 100% of synthetic X-rays. This performance significantly exceeded the results of board-certified radiologists and leading multimodal large language models (LLMs) featured in the original study conducted by the Icahn School of Medicine at Mount Sinai.
AI or Not achieved 100% detection of synthetic X-rays, compared to 41%–75% for human radiologists.
The platform maintained a 95% overall accuracy rate with a 7.8% false positive rate on authentic images.
General-purpose LLMs (including GPT-5 and Gemini 2.5 Pro) trailed with accuracy ranging from 57% to 85%.
The benchmark utilized an out-of-distribution dataset not present in AI or Not’s model training data.
Synthetic medical images are identified as high-risk factors for insurance fraud, legal tampering, and patient safety.
Researchers found no correlation between a radiologist’s years of experience and their ability to spot deepfakes.
The Mount Sinai study revealed a significant vulnerability in medical imaging, noting that even when radiologists were warned of the presence of AI-generated content, their accuracy peaked at approximately 75%. Purpose-built detection technology proved more reliable than clinical expertise alone, which is typically trained to recognize pathology rather than the subtle, "too perfect" artifacts of generative AI, such as unnatural symmetry or uniform soft-tissue textures.
"The Mount Sinai team showed just how vulnerable medical imaging is to generative AI, and we needed someone to put numbers on it," said Anatoly Kvitnitsky, CEO & Founder of AI or Not. "Our benchmark shows that purpose-built detection can close the gap that human experts and general-purpose AI models cannot. We're publishing these results to support the conversation, not to end it."
The rise of synthetic radiographs poses several systemic risks. Beyond clinical misdiagnosis, fabricated imaging can be used to support fraudulent insurance claims or introduced as tampered evidence in litigation and disability cases. Furthermore, synthetic images threaten research integrity by polluting training datasets used for future medical AI models. Protecting these workflows requires a multi-layered defense including clinician training, watermarking, and robust detection APIs.
AI or Not is the leading AI detection API for images, text, audio, video, and deepfakes. Powered by industry-leading models that deliver 98.9% accuracy, the API enables developers, businesses, and enterprises to embed synthetic media detection directly into their products and workflows. Trusted across a wide range of industries and use cases — from media and fintech to fraud prevention and identity verification — AI or Not is built for speed, precision, and scale.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.