FreeAIScannerProvenance
Check for Content Credentials and other signed origin information when present.
Evidence-led image review guide
How an AI Image Scanner Evaluates Evidence
Understand how image scanners evaluate provenance, metadata, and model signals associated with AI generation or editing. The live scanner is not currently available.
Educational content · No image upload · No live classifier
Evidence ledger
What the image scanner checks
An image can lose metadata when it is screenshotted, compressed, cropped, or reposted. A visual classifier can also be wrong. These guides keep different kinds of evidence separate.
File Metadata
Review software, encoding, dimensions, and modification clues without treating missing metadata as proof.
Model Inference
Use a visual classifier as one signal, with model information and limitations shown beside the result.
Process
How the AI image scan works
- 1
Upload or paste an image
Choose a supported image from your device or clipboard.
- 2
Inspect multiple evidence layers
Check available provenance, file metadata, and model-based visual signals.
- 3
Review evidence and uncertainty
Understand why a conclusion was reached and what cannot be determined.
Illustrative result anatomy
What a careful result should explain
This is an educational example, not an analysis of an uploaded image and not output from a live provider.
Observation
State what was actually found, including when evidence is missing or unavailable.
Interpretation
Explain what the observation may support without turning uncertainty into proof.
Limitation
Identify transformations, missing context, and other reasons the signal may be wrong.
Seven-state contract
Results designed for careful decisions
None of these states should be treated as proof of authorship, intent, fraud, or misconduct.
Direct AI provenance found
Recognized signed provenance or a generator declaration is present.
Likely AI-generated or AI-edited
Multiple available signals are consistent with AI generation or editing.
Mixed or uncertain
Available signals disagree or are too weak for a clear conclusion.
Likely camera or human origin
Signals lean toward camera or human origin, but AI involvement cannot be ruled out.
Insufficient evidence
The file does not contain enough reliable evidence for a useful conclusion.
Unsupported or unreadable
The file could not be safely parsed or is not supported.
Scan not completed
The scan ended before a result could be produced.
Interpret carefully
Why screenshots and edits are harder to assess
Screenshots replace the original file container. Cropping, recompression, and social platforms may remove signed provenance and camera metadata. A missing signal therefore cannot establish human origin, while model inference may be less reliable after transformation.
High-stakes boundary
Do not use a scan result as the sole basis for discipline, grading, employment, credit, housing, legal, or other high-stakes decisions.
Current availability
Learn now; upload later
The public site is intentionally content-only. No image upload, classifier request, account, payment, or advertising is active. Privacy-friendly aggregate analytics measures site usage without analytics cookies.
- Read the evidence methodology
- Understand common failure modes
- Review all seven result states
- Check the current privacy boundary
Questions
AI scanner questions
Can an AI scanner prove that an image was generated by AI?
No. A scanner can surface provenance, metadata, and model-based signals, but those signals may be incomplete or wrong. Treat the result as one part of a broader verification process.
Can I upload an image now?
No. The public site currently has no image upload or live classifier. The feature remains disabled while provider and data-handling terms are reviewed.
Does FreeAIScanner require an account?
No account is required to read the current educational site, and no account feature is active.
What does it mean if an image has no metadata?
Missing metadata does not mean an image is human-made. Social platforms, screenshots, editing tools, and compression can remove or change metadata.
Can the scanner detect AI-edited real photos?
It may surface signals associated with AI editing, but coverage varies by tool, edit, and file history. Review the evidence and limitations shown with each result.
Can I use a result to accuse someone of cheating or fraud?
No. The result is not proof of authorship, intent, cheating, or fraud. Use source verification and qualified human review, especially for consequential decisions.
Verify before you decide
Understand the evidence before trusting a result
Learn what provenance, metadata, and model inference can—and cannot—tell you.