A convincing photograph, video or voice recording is not proof that the event it appears to capture actually happened.
AI-generated media can look authentic. Yet genuine media can also look strange because of editing, poor lighting, compression and recording conditions. When authenticity matters, begin with the source, the original file if possible, and the circumstances of publication.
1. Preserve what you received
Keep the original file where possible. Record the URL, account, publication date, communications and when you first encountered it. Avoid re-saving or forwarding the only copy through services that might strip metadata. Save relevant surrounding messages and a dated note explaining what raised your concern.
2. Look for inconsistencies, not instant verdicts
Unusual reflections, lighting, shadows, object boundaries or speech timing can justify closer examination. But camera artefacts, normal image processing and compression can create similar effects. An odd-looking hand, a missing shadow or an abrupt audio pause does not, by itself, prove artificial generation.
3. Investigate provenance
Who originally published the content? Can the purported creator supply an earlier, complete source file? Does the claimed time or location match independent information? Are there other sources that corroborate the central claim? Source history often tells us more than the superficial appearance of an image.
4. Treat AI detector scores cautiously
Automated tools may help identify features worth exploring, but both false positives and false negatives are possible. A generic AI detection percentage should not be used as a definitive finding about authorship, the origin of a voice recording or the authenticity of a photograph.
5. Ask what the evidence establishes
Responsible findings separate observable facts, supported inferences, competing explanations and unresolved uncertainty. A useful report explains what additional evidence would help. Sometimes the sound conclusion is that the material cannot currently be authenticated.
Preserve the material and the relevant communication trail. Do not disclose passwords, transfer money or follow urgent instructions purely because a message or recording looks or sounds convincing. Independently verify sensitive requests using a trusted contact method.
How Verigate Verify approaches a case
We begin with the question you need answered, agree a proportionate scope, examine the available source information and supporting evidence, and explain our findings and their limits. We do not promise a guaranteed AI-versus-human verdict.
Discuss your material View example report (PDF)
General guidance, not legal advice or certification. Verigate Verify · October 2026.