Kapwing finds TikTok feeds carry three times YouTube’s AI slop rate
#AI

Kapwing finds TikTok feeds carry three times YouTube’s AI slop rate

Business Reporter
5 min read

Kapwing’s fresh-account test puts TikTok’s AI slop problem at 59%, with children’s videos showing the heaviest concentration.

Featured image

Kapwing counted AI slop in 294 of the first 500 videos TikTok showed a fresh account’s For You feed, a 59% rate that puts TikTok far ahead of YouTube in the company’s latest platform audit.

The video creation company Kapwing ran the same test on YouTube Shorts and counted 104 AI slop videos in the first 500 recommendations, or 21%. That gives TikTok about three times YouTube’s rate in a new-user feed, based on Kapwing’s test.

Kapwing also reviewed 10,742 TikTok videos across 20 categories. Its team checked featured videos tied to at least three popular tags per category, then grouped the results by subject area.

The company defined AI slop as videos with obvious AI-generated visuals, plus low-grade compilations built from AI scripts and voiceovers. That definition matters for marketers because it captures the common format now crowding short-form feeds: synthetic visuals, synthetic narration and fast assembly for volume.

Kids content leads the count

Kapwing found the highest concentration in children’s videos. In the Kids category, reviewers counted AI slop in 57% of 2,000 videos.

The tag-level numbers ran higher. Kapwing counted AI-generated videos in 97 of 100 featured videos under #cartoonkids. The #cartoons and #babysong tags each reached 83%, while #forkids came in at 79%.

Those numbers point to a production pattern. Creators can generate cartoon scenes, nursery-song clips and voiceovers at low cost, then publish many variants without filming children, sets or performers. TikTok’s recommendation system then decides which clips reach users.

Science and Education followed Kids at 35%. Health and History each reached 33%. Those categories lend themselves to voiceover explainers and image-based illustration, which lower the cost of synthetic production.

Fashion, Music and Fitness had the lowest rates in Kapwing’s review. Fashion came in at 1.3%, Music at 1.5% and Fitness at 1.6%. Those formats still favor on-camera presence, physical demonstration and recognizable performance.

TikTok faces a feed quality problem

TikTok has already moved to give users more control over AI-generated content. The company has tested AI content preferences inside its Manage topics settings, and it has worked on labels and invisible watermarks for AI media.

Kapwing’s findings suggest user controls have not solved the default new-user experience. A fresh account saw AI slop before it had history, follows or strong preference signals. That matters because TikTok uses the For You feed as the entry point for new users and as the main distribution channel for creators.

For brands, the issue goes beyond annoyance. Marketers that invest in TikTok now compete inside a feed where low-cost AI clips can outnumber human-made videos for new accounts. That changes creative benchmarks, testing cadence and brand safety reviews.

A brand cannot assume that polished human-shot content will stand out because it looks more expensive. In some categories, users may see a dense run of synthetic clips before a brand’s video appears. That context can make conventional creative feel slow, or it can help a credible human presence cut through the feed.

YouTube has taken a harder monetization line

Kapwing’s YouTube data gives TikTok a useful comparison. YouTube’s 21% Shorts rate still signals a large AI slop problem, but the platform has spent more time tying creator behavior to monetization rules.

YouTube CEO Neal Mohan has identified AI slop as a content quality issue, and YouTube has built detection systems around repetitive or low-value AI content. The company has also adjusted monetization policy for mass-produced and inauthentic material.

TikTok has emphasized labels and user controls. Those steps help users identify synthetic media, but they do less to change creator incentives if AI clips keep receiving distribution.

Kapwing said TikTok had labeled 1.3 billion videos as AI-generated by November. That number shows scale, but labels do not answer the core distribution question for advertisers: how much synthetic content does TikTok recommend before a user asks for less of it?

The market impact

Short-form video platforms now face a supply shock. Generative video tools cut production costs, and creators can turn one prompt into many clips. Platforms must rank that content against creator-made videos, brand campaigns and publisher work.

Advertisers care because feed composition affects attention and trust. If users associate a category with low-grade AI clips, brands in that category may need stronger proof signals: real people, clear sourcing, product demos and recognizable expertise.

The children’s content numbers raise a sharper concern. Parents, regulators and advertisers have spent years scrutinizing kids’ video recommendations on major platforms. Kapwing’s TikTok findings add AI-generated volume to that debate, with cartoons and nursery-song tags showing the heaviest concentration.

Kapwing’s study has limits. The 59% figure comes from one fresh-account test of 500 videos. A different account, location or viewing pattern could produce a different feed. Kapwing also sells video creation tools, so the company has a business interest in the line between human-made and AI-generated content.

Even with those limits, the category review gives the report weight. Kapwing’s team reviewed more than 10,000 TikTok videos, and the pattern across Kids, Science and Education, Health and History fits the economics of generative production.

Brand strategy shifts

Brands should treat AI slop as a competitive condition on TikTok. The feed no longer contains user-generated culture alone. It also contains synthetic content farms that can publish at a pace a brand team will not match.

That pushes brands toward proof and presence. On-camera experts, behind-the-scenes footage, product use and live responses can signal that a real organization stands behind the post. For educational or health content, brands should cite sources and show credentials inside the video, because synthetic explainers can mimic authority.

TikTok will need to show whether its controls change what new users see by default. YouTube will face the same test as its Shorts feed grows. For now, Kapwing’s numbers give advertisers a clearer benchmark: in one fresh-account test, TikTok showed AI slop in nearly six of every 10 For You videos.

Comments

Loading comments...