Slovenčina 
Čeština 
Shqiptare 
Österreich 
België 
български 
Bosanski 
Українська 
Eesti 
Македонски 
Türkçe 
Deutsch 
Dansk 
Español 
Français 
Suomi 
Ελληνικά 
Hrvatski 
Magyar 
Schweiz 
Italiano 
Lietuvių 
Latviešu 
Nederlands 
Norsk 
Polski 
Português 
Română 
Srpski 
Slovenščina 
Svenska 
TikTok has been working for some time to transparently label generated videos. In the coming weeks, it will introduce new tools that will fundamentally influence how AI-generated content spreads and how users will see it in their feed.
For e-commerce, this is an important development. TikTok is today one of the most powerful channels for product discovery, and AI tools have accelerated content production at a scale we never dreamed of.
AI Comes to Feed Personalisation
The new control will appear in the Manage Topics section, where users can already set what types of content they’re interested in. Now there will also be a setting for AI-generated content. However, it’s not a simple on or off switch. Rather, users will set whether they want to see more or fewer such videos. If someone enjoys AI content about history, TikTok will offer it to them more often. If someone doesn’t like AI, the algorithm will naturally push it aside.
TikTok isn’t the first to come up with the idea of limiting AI content display. Pinterest already offers a similar filter. The difference, however, is that TikTok has an incomparably more aggressive recommendation algorithm, and therefore a greater impact on what users see.
We can therefore expect that the reach of AI videos will be less predictable. Part of the audience will add them, another part will filter them out. We should be prepared for differences in campaign performance even with the same type of creative.
Sourc: TikTok
TikTok Strengthens AI Content Labeling Through Invisible Watermark
The second major innovation is invisible watermark technology that only TikTok can read. The platform already uses C2PA Content Credentials – a standard that embeds metadata about origin into images and videos. However, this standard has a weakness: when a video is downloaded, trimmed, edited, or re-uploaded, these data can technically be removed.
Invisible watermarking works differently. It’s not part of regular metadata and cannot be removed by editing the file. TikTok will add it to videos created with its own tools (such as AI Editor Pro) or to videos that already contain a C2PA identifier.
The goal is to tighten and automate AI content labeling. TikTok will no longer rely only on what the creator labels. If a video was created using AI, the platform will recognise it and add a label automatically.
In categories where customers need to see a real product and believe what they’re watching – for example, in cosmetics, health, food, or luxury goods – prominent AI content labelling can be detrimental and reduce credibility. That’s why it’s important to distinguish which types of videos to create using AI and which ones preferably not.
How AI Affects Value And Trust in Videos
TikTok reminds us in its report that AI is part of its creative and security tools, but equally important is building mechanisms that ensure transparent labelling of generated content. That’s why it’s now testing a hidden watermark and new options for regulating AI videos in recommendations.
In the broader context of platform development, we see that major players are experimenting with generated videos as a new format. At the same time, they’re beginning to address how AI content should be labelled and controlled.
This is a practical step that doesn’t speak to whether there should be more or less AI content, but rather that some platforms want to give users the option to decide for themselves.
This is a significantly different strategy than pushing AI content everywhere the technology enables its rapid production. E-commerce should take note as well.
What This Means for E-commerce and Marketing
- AI video performance won’t be uniform. The same type of content can work very well for one group of users and hardly at all for another. It will depend on how people set the level of AI in their feed and what type of content they prefer.
- Automatic AI content labelling will become increasingly precise. Platforms will be able to identify generated videos even without the creator labelling them. For brands, this means that even slightly modified or re-uploaded AI content will still be recognized.
- Creative work will have to split into two worlds. AI content is fast and efficient for testing topics or explaining complex things. However, real videos with people will remain key where trust is crucial. Keep this in mind, especially for content meant to help with the final purchase decision.
- User behaviour may change faster than we’re used to. If people start deliberately limiting or adding AI content, it will change the composition of their feed. This can affect which video formats will gain strength in the long term and which will start to weaken. Marketers will need to monitor whether their audience belongs more to the “AI-friendly” or “AI-averse” group.
