Glossary: Deepfake
Launch library · evergreen read
A deepfake is synthetic media, usually video, image or audio, generated or altered using machine learning to depict someone saying or doing something they never actually said or did in reality. The term blends 'deep learning' with 'fake', reflecting the underlying technology behind the technique itself and the deception it is often used to produce.
Quality varies enormously across examples. Early deepfakes often showed telltale flaws around eyes, teeth and lighting consistency, while more recent examples can be convincing enough that technical detection tools, rather than the naked eye alone, are increasingly the more reliable check available to anyone trying to tell the two apart.
Because detection keeps getting harder over time, source and context matter more than visual scrutiny alone ever could. Asking where a clip first appeared and whether any credible outlet has corroborated it is often a sturdier defence than trying to spot artefacts by eye alone.