35milimetre

AI influencer marketing has moved from a novelty to a small industry. A recent Mashable feature, Fake influencers, real money: Inside the AI creator business by Olivia Tauber, maps it clearly: platforms that generate faces and voices, agencies that manage fictional talent, and people selling courses on how to start. The people on screen do not exist. The money, and the people behind the accounts, do.

We wrote earlier about what actually goes into making an AI influencer work, mostly from the production side. This piece looks at the other side: how the business makes money, where it goes wrong, and what a brand should check before paying an AI creator or building its own.

How AI creators actually make money

Views alone do not pay much. Mashable's piece explains that an AI account earns in three main ways: platform payouts, brand deals and selling something to the audience.

Platform payouts have rules. According to the article, YouTube can pay AI-generated channels through its Partner Program, but it lists mass-produced AI content made with generic templates, without the creator's own perspective, as not eligible. TikTok's Creator Rewards Program pays for original videos longer than one minute. On TikTok and Instagram, an AI clip can reach millions of people and still earn almost nothing directly.

So many accounts use free videos as advertising. Some send viewers to paid subscription platforms. Some sell courses on how to build an AI character. And some sign brand deals: Mashable lists examples such as Lil Miquela appearing in a BMW campaign and imma appearing in an SK-II ad. There are also agencies that design and manage fictional talent, and companies running rosters of more than a hundred AI characters.

For a brand marketer, the takeaway is simple. An AI creator is a media channel run by someone. Before you buy space on it, you need to know who that someone is and how they built the audience.

Where it goes wrong

The Mashable article is honest about the problems, and three of them matter directly to brands.

Copied performances. Some tutorials recommend downloading a viral video and swapping the performer with an AI character, keeping the motion, framing and lighting. Mashable reports comedians finding their own jokes, gestures and outfits reused by AI accounts with large followings. A label saying "AI generated" does not fix this. As the article notes, a label does not show where the original performance came from or whether its creator agreed.

Engineered bodies. Many of these accounts use exaggerated, sexualised bodies to grab attention and push viewers toward paid content. A brand that appears next to that content takes on its reputation.

Fragile accounts. Mashable describes an AI comedian account that grew to hundreds of thousands of followers in less than a month, and then appeared to be deactivated. Fast growth built on borrowed material can disappear just as fast, along with any campaign you placed there.

None of this means AI characters are bad by nature. It means the method matters more than the follower count.

Disclosure is now the minimum, not the finish line

Platforms have added rules. According to the same article, Meta applies "AI info" labels, TikTok introduced an AI-labelling tool in 2023, and YouTube began rolling out its disclosure tool in 2024. TikTok also bans some synthetic depictions even when labelled, such as likenesses of minors and of private adults used without permission.

For brands, disclosure is the floor. Your audience should know when a person on screen is not real. But a disclosed character can still be built on someone else's work, and that is a risk no label covers.

What to check before you work with an AI creator

If an agency or creator offers you an AI influencer for a campaign, ask these questions. They are simple, and the answers tell you a lot.

Who runs the account? A named company or person you can sign a contract with.
Where does the movement and voice come from? Filmed by the operator, licensed, or taken from other people's videos? Ask to see a sample of the source material.
Who owns the character? If you build a campaign around the face, can it appear for a competitor next month?
How did the audience grow? Steady growth with consistent content, or a sudden spike from borrowed viral clips?
What else does the account sell? Look at the profile links. Your product will sit next to them.
Is it clearly disclosed? Check the bio and the platform labels, not only the contract.

If you are thinking of building your own character instead, the same questions apply in reverse. Use your own performers, or people who have agreed in writing. Keep a simple record of how each video was made. Design a character that fits your brand, not one designed only to win attention at any cost.

Our view

We have worked with AI imagery for about the last year, including character work, and Mehmet has 26 years in post-production. From where we sit, the interesting part of this trend is not the fake person. It is the craft and the honesty behind it. A character that holds the same face across a season, moves like a real performer and is built from material you own is a real production job. A character built from someone else's viral clip is a shortcut that can turn into a liability.

A practical takeaway

Treat an AI creator like any media partner: check who runs it, how the audience was built and what sits next to your product. Disclosure is necessary, but it is not proof of a clean process. Ask where the performance came from.