Poll: Should every AI generated image and video be labelled

[KnotKnull]

TL;DR

• AI tools can now create realistic images, video and audio in seconds, and many people struggle to tell them apart from real recordings
• The transparency rules in Article 50 of the EU AI Act began applying on 2 August 2026, including duties to disclose deepfakes and to mark AI generated content in machine readable ways
• China's rules on labelling AI generated content took effect on 1 September 2025, and technical standards such as C2PA Content Credentials aim to record where media comes from
• Research by Gordon Pennycook and colleagues in 2020 found an implied truth effect, where warning labels on some false content made unlabelled false content seem more believable
• This poll asks whether you think all AI generated media should carry a label, and why

The BIG question?

Scroll through any social feed and you will probably see images or videos made with AI. Some are obvious, some are art and some are designed to deceive. Should every piece of AI generated media carry a clear label? Or would universal labelling be impractical, unfair to creators or even counterproductive

We are asking because rules and technology are changing quickly. Governments, platforms and camera makers are all experimenting with labels, watermarks and provenance records. Your vote helps show what an everyday community thinks. The replies are a chance to explain where you would draw the line

What the rules say

The European Union's AI Act includes transparency rules in Article 50, which began applying on 2 August 2026. People who deploy AI to create deepfakes must disclose that the content is artificially generated or manipulated, with lighter requirements for evidently artistic, satirical or fictional works. Providers of AI systems that generate content must also mark outputs in a machine readable way so they can be detected. The European Commission has been developing a code of practice to help companies apply these rules

Other regions are moving too. China's measures for labelling AI generated synthetic content took effect on 1 September 2025, requiring both visible labels and hidden markers in many cases. In the technology industry, the Coalition for Content Provenance and Authenticity has developed C2PA Content Credentials, which attach tamper evident records showing how a file was created and edited. Several camera makers, editing tools and platforms now support them

Detection tools are part of the picture as well. Some AI companies embed invisible watermarks, such as Google DeepMind's SynthID, so their own outputs can later be identified. Independent detectors try to spot synthetic media without any watermark, though their accuracy varies and they can be fooled. No single tool can yet guarantee that a piece of content is real or fake

The case for labelling everything

Supporters argue that people have a right to know when what they see is synthetic. Labels help viewers judge content fairly and make deception harder. They also support trust in genuine photographs and recordings, which can otherwise be dismissed as fake. A simple universal rule is easier to understand than a patchwork of exceptions

There is a democratic argument as well. Elections, public health messages and breaking news are all vulnerable to convincing fakes. Consistent labelling could slow the spread of misleading content at critical moments. Many people say they would welcome that, even if labels are imperfect

Labels can also protect the people shown in synthetic media. Fake intimate images and impersonation videos cause serious harm to the individuals targeted, many of whom are not public figures. A clear marker that content is artificial can limit reputational damage. For victims, being able to point to a label or a provenance record can make a real difference

The case against

Critics point out that bad actors will not label their fakes, and hidden watermarks can sometimes be removed. A label requirement mainly burdens honest creators. AI is also used in ordinary editing, such as removing background noise or improving a photo's lighting, which makes it hard to decide what counts as AI generated. Labelling everything could make labels so common that people stop noticing them

Research adds a subtler concern. In 2020 Gordon Pennycook and colleagues reported an implied truth effect, where attaching warnings to some false headlines made unlabelled false headlines seem more accurate to readers. Applied to AI media, a missing label might be read as a sign that content is real. Labels may work best alongside other habits, such as checking sources and looking for provenance records

Questions to discuss

• Should labels apply to all AI use, or only to realistic images of real people and events
• Do you look for labels or provenance information today
• Who should be responsible for labelling: creators, platforms or AI companies
• Would labels change how much you trust what you see online

Closing thoughts

Labelling AI content is one of the most practical questions in technology policy right now, and it affects everyone who uses the internet. Rules are new and still being worked out, so public opinion really matters. Vote, explain your view and share any examples you have seen. Careful thinking here helps us all trust what deserves trust


Further reading

• Regulation (EU) 2024/1689, the Artificial Intelligence Act, Article 50
• Cyberspace Administration of China and others, Measures for Labeling AI-Generated Synthetic Content, effective 1 September 2025
• Coalition for Content Provenance and Authenticity, C2PA specification, c2pa.org
• Gordon Pennycook, Adam Bear, Evan Collins and David Rand, The Implied Truth Effect, Management Science, 2020