AI video is having its moment — and its authenticity problem

For years, AI-generated video was a curiosity — impressive for ten seconds, uncanny by the eleventh. That has changed. The outputs are now good enough to be genuinely useful for production, marketing, and prototyping, which is precisely the moment the interesting question stops being can it and becomes should we trust it.

The useful case is real

There is no need to be precious about this. Being able to generate and edit video from a description collapses work that used to require a crew, a location, and a budget. For storyboarding, for iterating on an idea, for producing variations of a concept before committing real money — this is a legitimate creative accelerant, and pretending otherwise helps no one.

The honest complication is that the same capability that makes it useful for a marketer makes it useful for a fraudster, and the two do not come separately.

Authenticity was doing quiet work

We rarely notice how much trust we placed in the sheer difficulty of faking a moving image. Video felt like evidence because, for most of its history, producing a convincing fake was expensive and slow. That friction was load-bearing. Remove it, and “I saw a video of it” stops being the settler of arguments it used to be.

The threat is less a single perfect deepfake than a world where any inconvenient real video can be waved away as a possible fake.

That second-order effect — the liar's dividend, where the mere existence of good fakes lets people dismiss authentic footage — may matter more than any individual forgery.

What actually helps

There is no clean technical fix, but the responses worth taking seriously cluster around provenance rather than detection. Trying to spot fakes after the fact is a losing race against improving generators. Establishing where a piece of media came from is more durable:

• Content credentials — cryptographically signed provenance attached at capture and carried through editing — shift the question from “is this fake?” to “can this prove where it came from?”.
• Disclosure norms — labelling synthetic media as a matter of course — are cheap, and brands that adopt them early bank trust they will be glad to have later.
• Institutional habits — newsrooms, platforms, and courts building verification into their workflows rather than trusting their eyes.

For marketers and communicators specifically, the practical takeaway is not to avoid the tools. It is to be the kind of organisation that discloses when it uses them, because in a world where seeing is no longer believing, a reputation for straight dealing becomes one of the few signals left that still carries weight.