Methodology
A signal is useful
when its limits stay visible.
This version uses an automated image-detection model to assess patterns associated with AI generation. The assessment can be wrong or inconclusive.
What is checked
One supported image is sent to the approved detection provider for analysis. The result is reduced to three user-facing findings: stronger signs of AI generation, no clear AI signs detected, or unable to reliably determine.
The service records the method version and checked time alongside the result. It does not claim to identify the generator, creator, source, or ownership.
What can change a finding
- Compression, screenshots, resizing, and editing can remove or add signals.
- Image quality, unfamiliar generators, and model coverage can affect reliability.
- False positives and false negatives are possible.
What a finding cannot establish
A finding does not prove that an image is real, fake, stolen, generated by a specific system, owned by a particular person, or connected to a genuine seller, payment, or event. Do not use it as the sole basis for a high-stakes decision.