TikTok serves nearly 60% AI slop to new accounts, Kapwing study finds

TikTok is showing brand-new users a feed that is, by Kapwing’s count, nearly 60% AI-generated junk.

In the Kapwing report, the team analyzed 10,742 TikTok videos across 20 popular categories and separately examined the first 500 videos shown on the For You page of freshly created accounts. Their central finding is simple, and uncomfortable: nearly six out of every ten videos served to a brand-new account are AI-generated junk.

Why this matters is not just “people are annoyed by spam.” It is about how TikTok builds trust with new users, and how quickly that trust can get corrupted. The For You page is the on-ramp. It is the algorithm’s first handshake. If the early experience is packed with AI slop, even high-intent users get trained on low-signal content before the system has enough behavioral data to find their actual interests.

There is also a brand and advertiser angle, and it is the one executives care about because it hits budgets. Brands typically do not buy “content quality” as a concept. They buy reach, engagement, and category relevance. If a large share of early impressions comes from AI-generated junk, the platform’s targeting and measurement can get shakier. The immediate consequence is wasted spend and weaker performance signals. The second-order consequence is harder: advertisers may demand tighter controls, more brand safety guarantees, or different attribution logic, all of which can change how TikTok monetizes.

Zoom out to incentives, because incentives are where algorithm problems often live. Content moderation and ranking have to balance user engagement, scale, and policy compliance. AI-generated junk is designed to pass as normal content at speed. The faster production gets, the more pressure the system faces to detect it early. And because Kapwing specifically looked at what new accounts see, this also hints at a key operational challenge: the platform has less history to lean on. With fresh accounts, there is no long behavioral trail to separate “curious early clicks” from “the user is being exposed to noise.”

Regulatory context is beginning to matter here, too. Even when regulation does not target TikTok directly, AI-generated media and deceptive content are increasingly in the crosshairs globally. The more platforms are expected to prevent misleading or low-quality automated content, the more executives will need to treat detection as a core product requirement rather than a behind-the-scenes policy task. A study like this is the kind of data points regulators and lawmakers can cite when they ask platforms what they are doing about synthetic content at scale.

There is a capstone implication for boards and investors: third-party measurement is becoming a lever. Kapwing’s analysis is not a screenshot taken from one account. It is a sample-based study grounded in specific counts: 10,742 videos across 20 categories, plus 500 early For You views. When that kind of methodology shows up in public, it pressures platform leadership to respond with either improved enforcement, clearer communication, or better measurement. If they do not, the narrative hardens, and “AI slop is rampant” becomes a brand risk and a political risk, not just a creator complaint.

For executives at other short-form and creator-driven platforms, this is a warning shot. TikTok is not the only place where AI junk can spread quickly, but the stakes are unusually high because the For You page is designed to discover new users fast. If the discovery layer gets polluted early, the entire flywheel can degrade: recommendations get noisier, creators face lower-quality incentives, and advertisers fight for clarity. Kapwing’s numbers, nearly 60% of videos shown to new users, are a reminder that the earliest moments on a platform are not “just onboarding.” They are the foundation of trust.