AI Casting
AI Avatars in Advertising: What Political Ads Teach Brands
US political campaigns have spent $137 million on AI-assisted ads. For brands, it is a case study in where a generated face works and where it puts trust at risk.
30 October 20264 min
An AI avatar in advertising can cost you your audience's trust in a single broadcast. Not because the video is badly made. Because nobody told viewers the face was generated, and nobody asked permission from the person it resembles.
US political campaigns are demonstrating this right now, at scale. For anyone commissioning a digital face, it is the most useful case study of the year. Our position is simple: a generated face works when it is disclosed, authorised and in service of a message. It breaks when it imitates a real person who never agreed.
Politics has already run the test at scale
In the 2026 midterm cycle, campaigns and groups have spent more than $137 million on ads made at least partly with AI tools. The Wesleyan Media Project identified 220 ads across 34 states. For the 2024 cycle it had found less than $50,000.
Only 27% of those ads carried a disclaimer. The count is probably an undercount: it is hard to tell when AI was used, and only 30 states require disclosure in political advertising.
For a brand, two things matter here. The technology is clearly mature enough for broadcast. And without internal rules, the decision about the face is left to whoever produces the ad.
A deceptive face burns trust, however realistic
The most discussed examples are not technical failures. They are polished ads showing scenes that never happened: a Senate candidate laughing at cards with extremist slogans, a senator shown at a beauty salon. As AFP reports, the second ad carried no AI disclosure, and the senator demanded its removal.
The public is not indifferent. In a survey cited by the same agency, 95% of respondents said it was unacceptable to create a video of an opponent saying something they never said.
In commercial advertising the mechanism is the same. A synthetic spokesperson who resembles a real person, with no consent and no disclosure, may expose the brand to complaints and takedown demands. Realism is not the problem. Deception is.
The damage reaches honest advertisers too
There is a less visible effect. Travis Ridout, co-director of the Wesleyan Media Project, warns in Fortune that voters may stop believing anything, and that some grow sceptical of candidates' credibility.
Translate that for a brand. When a category fills with undisclosed synthetic faces, even an honest ad is watched with suspicion. A brand that uses an avatar transparently is not immune to the context, but it holds one extra card: it can show how the work was made.
Casting a digital face is a brand decision
It is made before the first frame is generated. These are the criteria we use to assess an AI avatar for advertising:
- Identity: an original face with no reference to real people, or a real person with specific written consent.
- Scope of use: where, for how long, in which formats and for which products. Generic consent is not enough.
- Words and actions: the avatar says and does only what is in the approved script. No improvisation by the model.
- Disclosure: where and how the audience is told the face is generated.
- Consistency: the same face must hold across scenes, lighting and formats. This is not automatic and needs frame-by-frame checks.
An avatar does not replace a person in every situation. Where you need spontaneity, imperfection or a bond with a known person, a real face remains the better choice.
How we approach AI face casting
We start from the brief and one question: original face or real face? If it is a real person, we collect written consent first, covering scope, duration and formats, and only then start generating. We then build the face, test it across scenes and lighting, and review script and images with the client before delivery. The service is described on our AI Casting page. To see how it fits into a full spot, read our guide from brief to delivery.
What to ask a supplier before you sign
There is no need to distrust AI. You need to demand method. Before approving a digital face, ask:
- who the face is, or who it resembles, and where the written consent sits;
- which uses it is cleared for, and for how long;
- who approves every line and every scene;
- how AI use will be disclosed on your channels;
- how results are measured, trust included: see the ROI of AI content.
If a supplier cannot answer, the problem starts before the casting. To assess a face for your next campaign, start from the AI Casting page.
