Short clips on TikTok now include everything from dance trends and commentary to fully synthetic videos made with generative models. The platform has spent the last year building tools to identify and label this material, and those efforts have already led to more than 1.3 billion videos being tagged as AI generated.
The next step is giving users some control over how much of that synthetic content they actually see. TikTok is adding an AI specific control inside its existing “Manage topics” settings, combining user choice with stricter labeling and watermarking of AI generated clips. Together, these changes are an attempt to keep feeds usable when a rising share of content is no longer recorded directly from real life.
Why TikTok Needed AI Controls In The First Place
Generative tools make it easy to produce short, visually striking clips in large volumes. That convenience has filled TikTok with filters that alter faces, synthetic voiceovers, deepfake style edits, and fully generated scenes that only look like real footage.
Alongside that growth, TikTok has tightened its rules for synthetic media. The platform requires creators to label realistic AI generated images, audio, or video, particularly when they depict real people or could be confused with authentic recordings. Automated detection models and creator tags both feed into a labeling system that now covers more than 1.3 billion videos.
The scale of that number captures why an AI specific control has become necessary. For many viewers, a feed dominated by synthetic clips contributes to the same uneasy feeling described in discussion of a “dead” or artificial internet, where much of what appears in public feels generated or staged rather than genuinely social.
How The New TikTok AI Slider Works In “Manage Topics”
TikTok already allows people to adjust broad categories in their For You feed through a feature called “Manage topics”. Users can ask to see more or less of certain themes, such as sports, travel, current affairs, or comedy, and the recommendation system gradually reshapes what appears.
The new AI control works in a similar way, but it targets synthetic content specifically. Inside Settings, under Content preferences and then Manage topics, users will see a dedicated AI generated content slider. Moving it toward “less” signals that they want fewer AI generated videos in their For You feed. Moving it toward “more” does the opposite and invites the system to recommend more AI assisted clips.
This slider does not block AI content completely and it does not override all other ranking signals. Instead, it provides a clear preference that the recommendation system can weigh alongside other factors such as watch history, interactions, and regional trends. Over time, the goal is for feeds to reflect not only what people watch, but also how much synthetic media they are comfortable seeing.
Labeling And Watermarking: From Content Credentials To Invisible Marks
User facing controls only work if the platform can reliably tell which videos are AI generated. To support that, TikTok has been expanding its labeling and provenance tools on several fronts.
First, the service provides explicit labeling tools for creators and requires self disclosure when videos contain realistic synthetic elements. These labels appear as badges or notices on screen, helping viewers understand that what they are watching has been generated or altered by AI.
Second, TikTok has partnered with the Coalition for Content Provenance and Authenticity and adopted C2PA Content Credentials. This standard embeds metadata into media files so that platforms can recognise when a video has been marked as AI generated by other compliant tools. As a result, TikTok can automatically label imported clips that arrive with these credentials already attached, even if the uploader does not add a tag manually.
Third, TikTok is testing “invisible watermarking” for content created with its own AI tools and for material flagged through C2PA metadata. A recent TikTok newsroom update on AI generated content transparency describes this watermark as a non visible signal that remains readable by TikTok systems even after basic edits or reposting. The watermark is intended to support automatic labeling, provide a more durable provenance trail, and make it harder to strip away all evidence that a clip is synthetic.
Together, these systems create a layered approach. Visible labels tell viewers that AI was involved. Metadata and watermarking help TikTok decide when and how to show those labels, and they support future checks if a piece of content starts spreading more widely.
Will These Controls Make TikTok Feel Less Like A “Dead” Feed?
TikTok’s AI slider, labeling rules, and watermarking tests are part of a broader attempt to keep the platform usable when synthetic media is no longer a fringe category. They do not remove AI generated content, and they do not guarantee that every synthetic clip will be correctly tagged. They do, however, give users more say in how much machine generated material they see and provide regulators with clearer evidence that the company is treating transparency as a responsibility rather than a marketing line.
On their own, these measures will not resolve every concern about feeds that feel over automated or repetitive. The recommendation logic that decides which formats thrive, and the commercial incentives that reward constant posting, still shape how “alive” or “staged” the internet feels for regular users. A Dead Internet Theory explainer on ByAlexdavid argues that part of the discomfort comes from this broader environment, where bots, templates, and content mills narrow what surfaces in public.
Viewed in that light, TikTok’s AI features look like early infrastructure for a more honest feed. Labels, metadata, and watermarks cannot make synthetic media disappear, but they can make it visible as synthetic. Paired with controls that let users request more or less of this material, they at least move the platform away from a future where the synthetic and the real are indistinguishable by default.
