No defensible universal accuracy rate was retrieved from the available research links.
Benchmark conditions targeted compression, resizing, cropping, and multiple image generators.
Discovery results included Sightengine, Imagera, TruthScan, DeepAI, and Is It AI?
The search returned discovery links, not directly retrievable benchmark tables or complete official documentation.
Visual inspection alone is unreliable and can produce false positives, so no artifact should be treated as proof.
Create your account to keep this answer and continue from it later.
Let's look at alternatives: