It’s easy to guess that AI made a video of singing cats or that an AI chatbot is talking us through a customer service problem.

But we’re not as good at spotting AI as we think we are. Only half the time do we correctly guess that AI content isn’t made by a human, according to separateacademic studies.

Deepfakes – realistic AI-generated videos – can show politicians saying things they never said, helping them to win or lose an election. Eight out of 10 Europeans see AI content and other disinformation as a threat to democracy, according to a Eurobarometer survey in February and March.

Since yesterday (2 August), “Made by AI” labels are now a legal requirement across the EU, which will flag any newly AI-generated content.

Under the new rules, we'll also be warned before we interact with an AI service, or are about to be exposed to an AI system using biometric information to recognise our emotions or decide how to categorise us, such as by age or gender.

This is part of the EU’s AI Act, which requires businesses or websites targeting EU audiences to label AI-generated content. The label should be a machine-readable watermark – or even an AI icon. Companies are urged to add a text disclaimer somewhere. They can be fined up to €15 million, or 3% of their global revenue, if they don’t comply.

These rules don’t apply for personal use, so you can still make a clip of your cat meowing “Happy Birthday” to your grandma – although you might want to warn her that Mr Tibbs isn’t a real musical genius.

It’s still allowed to post artistic, satirical, or creative content online, just best to do it with a “made by AI” note.

There are more exceptions: you don’t need to flag that you’ve used AI for standard edits of a photo, such as removing a background or red-eye caused by a flash. You can also still run an AI grammar check on a document or colour-correct a video.

But businesses will need to alert users if they've used AI to insert new people into a photo or made “substantial” changes to their faces, or mimic someone's speech to say something they never really said.

The ultimate aim of these rules is to increase trust in what we read or see online, now that we can't always believe our own eyes. At best, it may help stop the floods of AI slop we're seeing. But these rules only cover the EU, and rely on people to be upfront about how they're using AI.