Marketing Content: Disclosure, Accuracy and Advertising Rules

AI can generate marketing content faster than your team can write it. A small business with limited marketing resources might be tempted to use AI to produce blog posts, social media content, product descriptions, and advertising copy. This is permissible, but it comes with obligations. The content must still be factually accurate, the AI-generated nature must be disclosed where required, and advertising claims must still be substantiated. Breaching these rules can result in complaints, takedowns, and in serious cases, regulatory fines.

Article 50 of the EU AI Act, in force since 2 August 2026, requires that content generated by AI be clearly marked in a machine-readable way. This is not optional for generative AI output used in marketing. If you use an AI tool to write a social media post, the platform you post it on must be able to read metadata indicating the content is AI-generated. Many social media platforms and content management systems are working to support this standard, but adoption is still early. For now, you can comply with Article 50 by including a human-readable disclosure near the AI-generated content, such as “This article was written with AI assistance” or “This image was generated using AI.” The machine-readable marking will likely become standard over the next two years.

Advertising standards in your jurisdiction still apply to AI-generated claims. In the UK, the Advertising Standards Authority requires that all advertising claims be truthful, that comparative advertising be fair, and that claims be substantiated by evidence. A language model can generate marketing claims that sound credible but are not true. This is called hallucination. You are responsible for checking whether the claims are true before you publish them. An insurance company cannot use AI to generate claims about coverage that are not in the policy. A fitness product company cannot use AI to generate health benefit claims that are not supported by evidence. If your marketing is flagged as non-compliant, you must be able to show that you checked the claims and that you have evidence backing them up.

For your small business, the practical approach is to treat AI-generated marketing content as a first draft. Use the AI to generate ideas and raw copy. Then have a human review it for accuracy, tone, and brand fit. Check any factual claims against your records. Check any medical, financial, or legal claims against your terms and conditions and any applicable regulations. Only after this review should the content be published. This review process is your defence if a complaint arises.

There is also the risk of accidental plagiarism or copyright infringement. Language models are trained on text from across the internet. When they generate content, they sometimes reproduce phrases or structures from their training data. If an AI tool generates marketing copy that happens to be very similar to a competitor’s copy or to existing published content, you are liable. The AI is not. You are. Check AI-generated marketing content for originality using a plagiarism detection tool before you publish it. This takes two minutes and protects you.

For images, the risks are different but equally real. AI image generation tools can create photorealistic images that look like they might depict real people, real products, or real events. If you use an AI-generated image in marketing and it looks like it shows a real person or real event, you need to disclose that it is AI-generated. This is required by Article 50. It also protects you from complaints that you are using deepfakes or misleading synthetic media. Some jurisdictions have specific rules about synthetic media in advertising. Disclosing that an image is AI-generated shows you are acting in good faith.

Stock images and visual content that you use to train or prompt AI image generators must be licensed for that use. Most stock image licences do not permit use in training AI models. If you use a licensed stock image as the basis for an AI generation, you may be breaching the licence. Check the terms before you do this. Similarly, if you use AI to generate a variation of an existing copyrighted work, you may be infringing copyright. The safest approach is to use only original content or content clearly licenced for this purpose as input to AI image generators.