If your “For You” feed has felt a little too synthetic lately—filled with uncanny deepfakes, AI-narrated stories that don’t quite add up, or visual “slop” that feels devoid of human touch—TikTok is working on a solution. The platform is currently testing a new control that puts the power back in your hands, allowing you to manually adjust how much AI-generated content (AIGC) you see.

But as with all things in the world of generative AI, the feature comes with a catch: it only works if the algorithm knows what AI is and what isn’t. To bridge that gap, TikTok is simultaneously rolling out invisible watermarking technology to better track synthetic media across its platform.

The new control is part of an expansion to TikTok’s “Manage topics” feature. If you aren’t familiar with it, this is the section of your settings (found under Content Preferences) where you can already tell the app you want to see less “Dance,” “Sports,” or “Current Affairs.”

In the coming weeks, a new slider labeled AI-generated content will appear in this menu.

Unlike a simple “on/off” switch, this functions more like a volume knob. Users can slide it to request “less” AI content, signaling the algorithm to down-weight those videos in the recommendation engine. Conversely, if you enjoy the new wave of AI art and storytelling, you can slide it to see “more.”

This nuance is important. TikTok isn’t banning AI; it’s treating it as a content category—a genre that some people love and others loathe. By integrating it into “Manage topics,” TikTok is acknowledging that for many users, AI fatigue is setting in, and they want their feed to return to human-centric creators.

The invisible problem

The effectiveness of this slider relies entirely on TikTok’s ability to identify which videos are actually made by AI. This is currently one of the hardest problems in the tech industry.

To tackle this, TikTok announced it is implementing invisible watermarks.

Currently, most platforms rely on metadata—specifically, a standard called C2PA Content Credentials. Think of this like a nutrition label embedded in a digital file that says “this was made with DALL-E” or “this was edited with AI.” While TikTok supports this standard, it has a major flaw: it is incredibly fragile. If a user screen-records a video or uses a tool to strip metadata before uploading, the “nutrition label” falls off, and the platform has no idea the content is synthetic.

TikTok’s invisible watermarks aim to solve this by embedding a signal directly into the pixels or audio of the content itself. This watermark is imperceptible to the human eye but readable by TikTok’s detection algorithms. The company says it will automatically apply these watermarks to:

• Content created using TikTok’s internal AI tools (like AI Editor Pro).
• Content uploaded that already has C2PA credentials attached.

This creates a persistent link. Even if the metadata is stripped later, the invisible watermark remains, allowing the algorithm to correctly categorize the video and respect your “see less AI” preference.

AI fatigue vs. AI integration

This move highlights a diverging path in how social media giants are handling the generative AI boom.

On one side, you have Meta (Instagram/Facebook), which is heavily integrating AI into the user experience, recently testing an “AI Studio” that lets creators make AI versions of themselves to talk to fans. On the other, TikTok is positioning AI as a variable that users should be able to dampen if it ruins their viewing experience.

This comes at a critical time. “Slop”—a slang term for low-quality, mass-produced AI content designed solely to farm views—has become a significant complaint among users across all social platforms. By offering a control mechanism, TikTok is trying to preserve the “authenticity” that made its platform famous in the first place, ensuring that human creators aren’t drowned out by machines unless the user explicitly asks for it.

What’s next?

TikTok has stated these changes will be rolling out over the “coming weeks.” It is important to note that this is a test. The slider will not be a magic wand that instantly removes every deepfake; it will only filter what TikTok’s systems can successfully detect. As AI tools get better at hiding their tracks, TikTok’s detection tools will have to improve to keep up.

For now, though, it’s a welcome step for anyone who just wants to see a real person dance, cook, or tell a story.