How to tell if an image is AI-generated
Use visual, file and provenance evidence together, then preserve an inconclusive result when the signals do not agree.
Read guide →Evidence-first learning
AI image authenticity guides
These guides answer common authenticity questions without treating one visual artifact, metadata field or model score as proof. Each guide separates observations from conclusions and states where verification must stop.
How to interpret AI detector scores
A detector score is a model output for a specific file and model version—not a probability of authorship or intent.
Read guide →Content Credentials explained
Learn what a valid, trusted credential can support and why a missing credential proves nothing about an image’s origin.
Read guide →Editorial standard
Every guide separates observation from conclusion.
Original file first. A screenshot or platform copy can remove provenance, rewrite metadata and change the pixels a detector sees.
Independent signals. Visual patterns, container evidence, signed provenance and source context answer different questions and should not be collapsed into one unexplained percentage.
Inconclusive is valid. When evidence conflicts or the evaluation scope does not match the file, the guide tells you where to stop.