Do I need to label AI-generated blog posts and marketing content?

This is the most common question we hear from marketing and content teams, and the honest answer is nuanced: ordinary marketing copy usually does not need a visible “written by AI” label — but two other duties probably do apply to you. The distinctions live in Article 50(2) and 50(4) of Regulation (EU) 2024/1689, both applying from 2 August 2026.

Duty 1 — Machine-readable marking (Article 50(2), providers)

Providers of AI systems that generate synthetic audio, image, video or text must ensure outputs are “marked in a machine-readable format and detectable as artificially generated or manipulated”. This is about metadata and watermarking — think C2PA Content Credentials — not a visible caption. If you build an AI writing or image tool, this is your duty. If you merely use ChatGPT, Midjourney or similar, the marking duty sits with those providers, not with your marketing team.

Exemption worth knowing: systems performing an assistive function for standard editing, or that do not substantially alter the input, are out — AI grammar fixes on human-written text are not “synthetic content”.

Duty 2 — Visible disclosure for two content types (Article 50(4), deployers)

Deployers — i.e. the business publishing the content — must disclose:

• Deep fakes: AI-generated or manipulated image, audio or video that appreciably resembles real persons, places or events and would falsely appear authentic (Article 3(60)). A photoreal AI “customer” in your ad, an AI voice-over cloned from a real voice, a fabricated “demo screenshot” — disclose it.
• AI text informing the public on matters of public interest: AI-generated news-style articles must be disclosed as AI-generated — unless the text “has undergone a process of human review or editorial control and a natural or legal person holds editorial responsibility”.

So where does a typical AI-assisted blog post land?

• Product pages, ads, landing copy drafted with AI, edited by your team: not public-interest text, not a deep fake — no AI Act labelling duty. (Consumer-protection law still forbids misleading claims, AI-written or not.)
• SEO articles published at scale with no human review: risky. If they read as informational journalism-style content, the Article 50(4) text duty can apply, and you can't claim the editorial-control exemption you didn't exercise.
• AI-generated “testimonials”, fake reviews, synthetic spokespeople:deep-fake territory plus misleading-practice territory. Disclose — or better, don't.
• Company newsroom / thought-leadership written with AI, with a named editor: the human-review exemption is available — document who reviewed and approved each piece.

How to disclose when you must

Article 50(5) requires the information to be clear and distinguishable at the latest at first exposure, and accessible. For published content, that means a caption or adjacent note (“This video was generated with AI”), not a line in your privacy policy. For evidently artistic or satirical works, the disclosure may be done in a way that does not hamper enjoyment of the work — a credits mention can suffice.

What content teams should set up before 2 August 2026

1. An editorial-review record for AI-assisted articles: reviewer, date, sign-off. This is what makes the Article 50(4) exemption stick.
2. A deep-fake rule: any photoreal synthetic person, voice or event in marketing assets gets a visible label, no exceptions.
3. Vendor checks: if you build AI content features into your own product, you inherit provider duties — see the full compliance checklist and, for non-EU teams, whether the Act reaches you at all.
4. Unsure which side of the line your content falls on? The free checker walks through these questions in about 3 minutes.