Expert Mode with Donatas Smailys, Co-Founder and CEO of Billo
What happens to a performance strategy the moment your audience learns to spot the trick?
For about two years, the advice handed to advertisers was blunt: lean into AI creators. They’re cheap, they never sleep, and you can spin up a hundred variations before lunch. That advice built real campaigns and, for a stretch, real returns. Donatas Smailys, co-founder and CEO of Billo, a user-generated content platform that turns creator videos into ads for Meta, TikTok, and YouTube Shorts, thinks the ground under that advice is moving. Just not the way the headlines say it is.
“I’d say it’s being corrected, not reversed,” Smailys says. The math hasn’t changed — AI is still cheap, and no one can argue otherwise. What’s changed is who’s watching. Synthetic content performed when it was new and nobody was looking for it. “That window is closed,” he says. Viewers have developed an instinct for synthetic faces, feeds are saturated with them, and TikTok is now teaching its own users how to catch AI, through a media-literacy guide built with outside experts.
So who’s actually exposed? In Smailys’s read, the brands most at risk are the ones using a synthetic spokesperson as the face of a performance campaign — especially when that spokesperson isn’t disclosed. “If a business depends on an AI person,” he says, “at minimum they risk being pushed out of recommendations.” The legal exposure stacks on top of that. New York’s synthetic-performer disclosure rules are already in force, and the EU AI Act‘s transparency obligations arrive on August 2. Then there’s the quieter cost: an audience that stops trusting you.
Here’s where Smailys pushes back on the coverage. TikTok’s actual move is narrower than the alarm around it. The platform is testing account-level detection aimed at AI spam in three high-risk buckets — politics and current events, financial advice, and medical content — while saying it still welcomes creators who use AI for original work. “If you’re a brand using AI somewhere in your workflow and labeling what needs to be labeled, nothing changes for you today,” he says.
But he doesn’t read that as a reason to relax. Platforms tend to start enforcement where the damage is worst — scams, fake medical advice — and widen the net once the detection holds up. TikTok has now said out loud that AI spam is crowding out real creators, and it’s building the tooling to catch that content at scale. “Brands that build around real people now won’t have to scramble when the scope widens,” Smailys says.
That distinction matters more than the human-versus-AI framing everyone reaches for. TikTok is, after all, selling AI ad tools and labeling AI videos rather than banning them — billions of them. So the real fault line isn’t human or machine. It’s authored versus spam. Smailys agrees, and then adds a second line most people skip.
“Billo is not anti-AI, we use it every day,” he says — to read performance data, to brief creators, to make the work faster. “What we don’t do is synthesize the person on camera, because we believe that real humans perform better.” He won’t pretend the platforms aren’t going the other way. TikTok’s Symphony suite will happily generate a presenter for you. But look at how TikTok treats its own product: the stock avatars are built from licensed, paid actors, and every Symphony video carries an automatic AI label. “Even the platform that sells avatars won’t let them pass as real humans,” he says. His conclusion: “The line is ‘authored vs. spam,’ but there’s a second line, disclosed vs. undisclosed, and platforms are enforcing both.”
On the regulation, Smailys is unusually complimentary. New York’s law, he argues, drew the line in exactly the right place. It doesn’t care whether AI was used to edit, caption, translate, resize, or test variations of a real creator’s ad. It triggers when AI generates the human on screen. “That’s the right place to put it,” he says, “and brands should adopt it as their internal standard everywhere, not just for New York.” The penalties run from $1,000 for a first violation to $5,000 for each one after — real money, but not the number that should scare anyone. “Even $5,000 per violation is nothing,” he says, “next to what happens when your audience realizes the person recommending your product doesn’t exist.”
It’s a clean articulation of the phrase he’s built Billo’s repositioning around: AI is not a face, it’s a tool. And it leads straight into the practical question a lean marketing team is actually asking — what do I do this quarter?
His answer starts with the law, not the strategy. Go through every live creative and flag anything where a person on screen was AI-generated, background people included, since those need disclosing too. Then ask your vendors and agencies, in writing, whether any AI-generated humans made it into what they delivered. Only after that does the creative decision come in. Point AI at the work it’s good at: editing, captions, testing, analysis. Take it off the front of the camera. He’d treat AI voices the same way, even where the law hasn’t caught up to them yet, because platforms label them and audiences notice. Whatever budget that frees up, he’d move into fewer real creators, tested properly.
None of this reads as nostalgia for a pre-AI creator economy. Smailys is running a company that leans on AI hard — just not at the one spot where, in his telling, the trust breaks. The bet underneath Billo’s whole position is that in a feed where anything can look real, the thing worth paying for is the thing you can prove is. For marketing leaders weighing how far to push automation into their creative, that’s the useful reframe: not whether to use AI, but where putting it will cost you the audience you were trying to reach.
