Meta is now labelling content as AI-generated, with help from AI vendors
In line with the EU AI transparency laws that came into effect on 2nd August 2026 [1], Meta is now identifying and labelling AI-generated content in their products. AI vendors are tagging anything made with AI-generator tools or responses, but it’s unseen to the user.
Whilst you can manually mark your own content as AI-generated, platforms are now able to do it on your behalf too.
Why is this happening?
Under the EU AI Act, it’s now a legal requirement to show transparency in how businesses are using AI. Companies must declare exactly how they are using AI and for what purpose/s.
This requirement now extends to transparency around content that we consume online. Meta has made the update to label content as AI-generated [2] based on its decision whenever it identifies that AI has been used in its creation, either in Meta or external products.
How does it decide?
Right now there are disputes as to what classifies as AI-generated.
For example, some creators have been caught out simply for using basic tools like the background remover feature [3] (I’m a fiend for using that).
Others have had their videos labelled for making use of automated caption creation. Again, I use this, so am left with a decision of priority - continue to make my videos accessible with closed captions and accept the AI-generated label on them?
How does it apply the label?
An interesting thing about this is that AI vendors are recording the activity secretly using metadata and watermarks without it being visible to the user.
Anthropic are applying invisible watermarks on AI-generated text [4] and OpenAI uses watermarks and C2PA content credentials [5].
These indicators are getting scanned and picked up by Meta within its products and the user could be none the wiser.
What does having the label on my content mean?
Technically speaking, having the AI-generated label is only a disclosure and does not reduce your reach for the content or intend to cause you any kind of account violations. The only thing that puts you at risk is usual content flags (such as misleading ads, impersonation of people etc.) or failure to disclose use of AI. [6]
However, creators could find this an unwanted feature, feeling fearful of being branded as part of the ‘AI slop’ [7] we see so much of around today.
Is this a good thing?
I have a few thoughts on this:
You don’t need a label to identify a lot of content out there right now as being visibly made with AI. For example, my partner and I took a simple walk down out local high street and there were multiple shop windows with AI-generated posters hanging up in them. You can see it from a mile off.
I agree with having transparency on when AI is being used, particularly from a customer perspective. If you have paid someone for time and effort in copywriting for example, to then find out that they generated a page of AI slop in seconds then that’s not okay.
Finally it is early days but it will be interesting to see updates on the sensitivity range of what gets picked up and flagged as AI-generated. As an example, generating captions on a video has improved efficiency but not generated any new, made up content so I do dispute this.
Let me know how you’re getting on with this new feature!
For any help with content strategy and creation, take a look at our Social Media Management services.