Deepfakes
Media content produced with artificial intelligence in which faces, voices or entire scenes are manipulated to look and sound convincingly real.
Also known as: AI Fakes, Synthetic Media
Deepfakes are image, video or audio recordings that have been generated or altered by generative AI models. The result feels deceptively real: a celebrity says lines they never spoke, a voice recites a text that was never recorded. Within a few years the technology has moved from research labs into consumer tools.
How deepfakes are made
Behind the scenes sit neural networks, often generative adversarial networks or diffusion-based models. They learn from real recordings how a person looks, how a voice sounds and how a face moves. With the trained model you can then produce any new content in which the person does or says things that never happened.
Where to take care
In marketing deepfakes appear as both a risk and a tool. As a risk: scam schemes, identity theft, reputational damage, fake testimonials. As a tool: localised ads without a reshoot, voice synthesis for service bots, virtual brand ambassadors. In both cases consent, legal admissibility and transparent labelling are mandatory, since GDPR and personality rights set clear limits.
Day to day marketing practice
Serious advertisers should check each synthetic person, each cloned celebrity and each AI voice up front. Training material must come from permitted sources, the consent of the person shown must be on record, and the labelling as AI generated content must be visible. Anyone using deepfakes in a creative should also think about audience perception, because once an audience notices a manipulation, trust evaporates quickly.