Skip to content
AI · MEDIA · ONLINE DECEPTIONEvidence before belief
← All examples

Audio & video · Worked scenario

That sounds like them. Is it?

Voice familiarity is a starting point, not a recording's chain of evidence.

Invented example for learning

A ten-second audio clip is shared as proof that a public speaker made a surprising statement. There is no full recording, publication date or original uploader. Several replies call it a voice clone.

What is actually known?

You have a short sound recording and an attribution. You do not yet know whether it is complete, edited, impersonated or generated. The voice-clone accusation is not established either.

How to investigate it

  1. Find the earliest attributable upload and request the full recording with its date and location.
  2. Compare the excerpt with a complete original recording if one is available, including the words immediately before and after it.
  3. Seek confirmation from an attributable source with direct knowledge of the recording.
  4. Treat unusual cadence, audio glitches and detector scores as supporting clues with alternative explanations.
  5. Keep an accurate transcript separate from the question of who spoke and how the audio was produced.

Where the evidence stops

A recognisable-sounding voice is not sufficient proof of identity or AI generation. When provenance is missing, describe that limitation rather than issuing a definitive label. This example is educational; the checker does not offer standalone audio-file uploads.

This is an illustrative scenario, not a report about a real incident or a published verdict. No media, source record or detector result has been invented as evidence.

Take the next step

Verify a video before sharing →

Check something of your own ↗Our methodology