AI tools can produce images, video and voices that look and sound real. That creates obvious risks: fake events, impersonation and misleading product claims. Platforms have responded with a mix of automatic detection, creator disclosure tools and policies on synthetic media. The Academy's glossary explains the CAP Code and regulated claims.
How platforms approach AI labels
| Platform | Approach |
|---|---|
| Instagram and Facebook | Labels such as AI info, applied when detected through industry standard signals or when users disclose; disclosure expected for realistic synthetic video and audio |
| TikTok | Requires labelling of realistic AI-generated content and can apply labels automatically using content credentials |
| YouTube | Requires creators to disclose realistic altered or synthetic content, with labels shown to viewers |
| Shows content credentials on some images where available |
Policies evolve, so check each platform's current rules before posting AI content.
Automatic detection
Many AI tools embed metadata or watermarks, such as content credentials, that platforms can read. If you upload an image made or edited with these tools, a label may appear even if you did not add one. Heavy editing in AI-enabled apps can also trigger labels.
When to disclose
- Realistic people, places or events that did not happen.
- Synthetic voices or avatars presented as real people.
- Altered footage that changes what actually occurred.
- Product images that could mislead about appearance or results.
Minor edits such as colour correction, cropping or removing a stray object generally do not need disclosure, although platform rules vary.
UK advertising rules
There is no specific UK law requiring AI labels on all marketing, but the CAP Code prohibits misleading advertising, whatever tools created it. An AI image that exaggerates a product's appearance or results is misleading. The ASA has noted that the same rules apply to AI-generated ads. The Academy's guide to AI images for social media covers safe uses.
Content reaching the EU
Brands whose content reaches audiences in the EU should be aware that the EU AI Act includes transparency obligations for certain AI-generated content, such as deepfakes. Check whether these apply to your activity.
Labels and trust
Audiences generally react better to honest disclosure than to discovering undisclosed AI later. A simple note, such as illustration created with AI, prevents confusion. For creative brands, explaining how AI fits your process can even build interest.
Labels on ads
Ad platforms have their own rules for AI in ads, including disclosure requirements for political and social issue ads that use realistic synthetic content. Check ad policies before running AI-generated creative.
Creators and AI
Creators using AI in sponsored content should follow platform disclosure rules and the brand's guidelines. Brands should include AI expectations in briefs and contracts. The Academy's guide on how to write an influencer brief covers briefing.
Impersonation and deepfakes
Never create AI content that impersonates real people without consent. Brands are also targets of impersonation through fake ads and deepfakes; monitor for misuse of your brand and report it quickly. The Academy's guide to social media scams targeting businesses explains how.
Building a disclosure policy
Decide which uses of AI your team discloses, how labels are worded and who checks. Include this in your wider AI policy. The Academy's guide to an AI policy for marketing teams explains what else to cover.
Keeping records of AI use
Note which tools were used to create or edit each campaign asset.
Labels in Stories and short video
For Stories, Reels and TikToks, place an on-screen note in the first frame where viewers will notice it, not just in the caption.
Wording examples
Short, plain labels work best: made with AI, AI-generated illustration, voice created with AI or edited with AI tools. Put the label where people will see it, such as on the image itself or at the start of a caption, rather than in a hidden comment. Instagram captions allow 2,200 characters✓, but the label belongs in the opening line.
Labels for text
Platforms generally do not require labels for AI-assisted captions or posts, and readers rarely expect them. Disclosure becomes important when text presents itself as a personal experience, testimonial or review that did not happen, because that would mislead. Never use AI to write fake reviews or customer stories.
When a platform labels content you did not expect
Sometimes platforms label content that was only lightly edited, because an editing app embedded AI metadata. If a label seems inaccurate, check the platform's process for reviewing labels, and consider whether the editing tool adds content credentials by default. A brief explanation in comments can reassure followers.
A worked example
A furniture retailer uses AI to create lifestyle scenes showing its sofas in imaginary living rooms. The sofas themselves are real product photos placed into AI backgrounds. The team adds a short caption note, room setting created with AI, ensures the sofa's colour and size are accurate and uses platform labels where required. Comments focus on the products rather than questioning the images.
Common mistakes
- Realistic AI people or events without disclosure.
- AI product images that misrepresent.
- Assuming no labels are needed because no UK law requires them.
- Removing metadata to avoid labels.
- Impersonating real people.
Learning AI responsibly
The Academy's AI tools module in the second level covers disclosure and safe use, and the legal and ethical module covers advertising rules. The free assessment shows where you would start.
