Newsroom · Eterna-owned guide
What Is a Deepfake and Why Is It Becoming a Reputation Problem?
A deepfake is AI-generated media depicting a real person doing or saying something they didn't. Here's how it works, and why it's become a reputation risk.
Synthetic media, defined
A clip shows up in your feed: someone you recognize, saying something that doesn't sound like them. The lighting's a little off. The voice is close but not quite right. Or maybe it's flawless, and you'd never have known if a caption hadn't told you. Either way, you're looking at a deepfake, and the term gets used for such a wide range of things that it's worth being precise about what it actually means.
What a deepfake actually is
A deepfake is media, video, audio, or an image, generated or altered using machine learning to depict a real person doing or saying something they didn't actually do or say. The "deep" refers to deep learning, the branch of AI used to generate the content; the term itself is a blend of "deep learning" and "fake," first popularized around 2017.
What makes a deepfake specifically different from ordinary photo or video editing is how it's made. Traditional editing tools, cropping, splicing, filters, are applied by a person, deliberately, frame by frame or effect by effect. A deepfake is generated by a model trained on examples of a real person's face, voice or likeness, which then produces new synthetic content built from patterns it learned, not a direct manipulation of an original file. That distinction matters because it changes what's possible: a well-trained model can generate entirely new footage of someone saying something that was never recorded in any form, not just alter something that was.
Why this has become a reputation problem, not just a technology story
Deepfakes get covered constantly as a technology story, how realistic the tools have become, how cheap they are to use. What gets less attention is the specific way this maps onto reputation risk, which comes down to three things happening at once.
- It requires very little source material. Generating a convincing likeness once required substantial footage and technical skill. That barrier has dropped considerably, a handful of public photos or a short clip of someone's voice from a video call or a conference talk can now be enough starting material. Anyone with any public presence, which, at this point, includes most professionals, not just celebrities, has that much material already available.
- It's hard to visually catch, even for people paying close attention. Research on human ability to spot high-quality synthetic video has found accuracy rates well below what people assume they'd achieve. That's not a criticism of anyone's attentiveness; it's a reflection of how far the underlying generation quality has advanced. Automated detection tools face a related problem: their accuracy, measured under controlled lab conditions, tends to drop substantially once applied to real-world content that's been compressed, reshared or altered from its original format.
- It arrives already carrying the weight of "seeing is believing." Video and audio have historically been treated as stronger evidence than text , people trust what they see and hear more readily than what they merely read. Deepfakes exploit exactly that trust, which is why a fabricated video tends to do more reputational damage, faster, than an equivalent false claim made only in writing.
Together, these three factors explain why deepfakes sit at the center of the modern reputation-risk conversation: the barrier to creating one has dropped, the barrier to spotting one has risen, and the format itself carries more persuasive weight than most other kinds of false content.
Who actually gets targeted
The popular image of deepfake targets is limited to celebrities and politicians, but the pattern is broader than that. Executives get impersonated in fraud schemes. Ordinary individuals, including students, get targeted in harassment, bullying and extortion contexts that never make national news but are just as serious to the people involved. Anyone with enough public photos or video to serve as training material is a plausible target, which in practice is nearly everyone with any online presence at all.
How verification actually works, and why "looks real" isn't the standard
Because visual inspection is unreliable, credible organizations working in this space don't rely on "does it look fake" as a verification method. Eterna applies a four-part standard before treating any finding as confirmed: the content must be traceable to a checkable source, corroborated by an independent second signal, attributable to a documented verification method, and, before any response is taken, authorized by the affected person. The Deepfake Verification Guide covers this standard in full.
What to do if this applies to you right now
If you're reading this because you've actually found a deepfake of yourself, Someone Made a Deepfake of Me, What Should I Do? has the immediate steps. This article is meant as background, useful before or after an incident, but not a substitute for that more direct guide if you need it right now.
FAQ
Is every manipulated video a deepfake? No. "Deepfake" specifically refers to AI-generated synthetic media depicting a real person. Traditional editing, cuts, splices, filters, slowed or sped-up footage, is manipulated but not a deepfake in the technical sense, even though both can be misleading.
Can deepfakes be audio-only? Yes. Voice cloning, which recreates someone's voice from a sample of real audio, is a form of deepfake and is increasingly used in impersonation and fraud attempts, separate from any video component.
Are deepfakes illegal? It depends on jurisdiction and how the content is used. Laws vary significantly by location and continue to evolve; this article is general information, not legal advice.
Is there a way to reliably detect deepfakes with software? Automated detection tools exist and continue to improve, but their accuracy tends to drop meaningfully outside controlled testing conditions.
Related reading
Someone Made a Deepfake of Me, What Should I Do?
The immediate, ordered steps if this is happening to you right now.
The Deepfake Verification Guide
The four-part standard behind calling an incident confirmed.
How AI Impersonation Can Damage Personal and Business Reputation
What happens once a convincing fake actually starts to spread.
Deepfake Protection: Detection, Verification & Response
How Eterna's protection framework applies once a deepfake is suspected.