Image, audio, video, or other media substantially generated or transformed through computational systems, including but not limited to deceptive uses.
Follow material in an order that builds understanding.
inAiLoading page content...
Follow material in an order that builds understanding.
inAiLoading page content...
Unit 03 · media evidence
Plan for this page
After this unit, you should be able to
Terms you will use
These definitions prepare you for the reading; you do not need to memorize them.Image, audio, video, or other media substantially generated or transformed through computational systems, including but not limited to deceptive uses.
Synthetic or manipulated media that convincingly represents a person, event, or identity in a way that can mislead about what occurred.
Machine-readable provenance information that can record origin and transformations when participating tools preserve and verify it.
Generated media can be realistic, and authentic media can look strange after compression, cropping, lighting, or editing. Searching for unusual fingers, blinking, shadows, or audio rhythm may suggest questions, but no permanent list of visual defects can serve as a truth detector.
Start with the claim around the media. Who first published it? When and where was it supposedly recorded? Is the account in a position to know? Does the content match known chronology, geography, weather, language, and surrounding events?
Find the earliest available version, reverse-search key frames or images, and seek independent reporting or primary records from the place and time. Contact the depicted person or responsible institution when appropriate. A derivative repost without context is weaker than an attributable original plus corroboration.
Content credentials can make origin and editing history more inspectable when the chain is present and valid. They do not prove that a staged scene was truthful. Missing credentials also do not prove that media is fake, because many cameras, editors, platforms, and historical files do not preserve them.
Urgent outrage creates pressure to share first. Pause consequential action, save the source and timestamp without redistributing sensitive content, and describe uncertainty precisely. Ask verification organizations or qualified forensic specialists when the stakes justify it.
Be alert to the liar’s dividend: growing awareness of deepfakes can let someone dismiss authentic evidence merely by calling it AI-generated. Apply the same evidence process to claims of authenticity and claims of fabrication. Avoid treating detection software as an infallible judge.
Interactive model
Record the exact allegation, earliest known source, date, place, uploader, and chain of reposts.
Misconception to correct
Active checks
Explain back and transfer
Source context
Defines technical standards for recording content provenance and edits; availability and valid credentials do not establish that the depicted claim is true, and absent metadata does not prove fabrication.
Supports managing synthetic-content, information-integrity, provenance, disclosure, testing, and monitoring risks; techniques and threat conditions continue to change.
Supports critical, ethical, human-centred evaluation of AI-produced information; it is not a forensic verification manual.