AI content labelling in 2026: what the EU AI Act means for your social media

Over the summer, AI labelling stopped being optional in Europe.

The EU AI Act’s code of practice on transparency for AI-generated content came into force at the start of August 2026. The key rules:

• Content generated or modified with AI must be labelled, in a way computer systems can detect.
• Chatbots and other interactive AI systems must tell people they’re talking to AI, not a human.
• Deepfakes (AI-generated or edited images, video and audio) must be labelled.
• New AI models launched in the EU must support machine-readable markings from day one.

Existing models get a transition period. Anthropic, for example, is adding invisible text watermarks across Claude. They don’t change the quality or meaning of the output, but they travel with the text when it’s copied and can survive light editing. If you’re using AI at work, assume there’ll always be a trace.

The UK isn’t in the EU, but we think it’s very likely to follow, as it did with GDPR.

You may already be seeing more “AI info” labels on Instagram. Canva now adds metadata labels to content made or substantially edited with its generative AI features. Canva has confirmed that minor tools, such as the background remover, don’t trigger a label. Features like background generator, Magic Expand and Magic Edit are more likely to. As a rule of thumb, if you see the sparkle icon in Canva, it’s AI. Some elements in Canva’s own library are AI-made too, and you can check by right-clicking and viewing the info.

Some critics are furious about all this. We think it’s a good thing. Influencer posts still work now they’re labelled as ads, and we expect AI content to go the same way. Being honest with your audience only gets more important from here.