EU AI Act from 2 August 2026: AI Marketing Texts Are Exempt From the Labelling Duty

From 2 August 2026, the transparency obligations under Article 50 of the EU AI Act apply. Ever since, one unresolved question has circulated in agencies and marketing teams: will AI-generated text have to be labelled as such? The debate was fuelled, among other things, by a widely shared LinkedIn thread and by the draft Commission guidelines that the EU opened for consultation on 8 May 2026. The sober answer: for ordinary marketing and SEO text, the disclosure duty under Article 50(4) does not apply. It is drawn more narrowly than the headlines suggest.

What applied before

Until the cut-off date, Article 50 was law on paper without application. The obligation had stood in the text since the AI Act was adopted, but it was not yet in force. In practice, anyone producing and publishing AI text did not have to think about a labelling duty derived from the AI Act. Requirements from general competition, copyright or data-protection law remained untouched, but a specific AI disclosure duty did not exist as an enforceable norm.

On top of that came uncertainty about scope. The statute refers to „text which is published with the purpose of informing the public on matters of public interest” — a formula that is broadly interpretable without guidance. It is precisely this gap that the Commission guidelines, in draft since May 2026, are meant to close.

What applies now

1. The duty is in force — but narrowly framed. Since 2 August 2026, deployers must disclose that a text has been artificially generated or manipulated — but only where that text is published „with the purpose of informing the public on matters of public interest”. The wording of Article 50(4) is the decisive filter here: not every published AI text triggers the duty, only the one with a public-interest connection.

2. Marketing and SEO sit below the threshold. The draft guidelines draw the line explicitly: advertising copy, product literature and internal documentation do not count as informing the public on matters of public interest. Political, social, economic, cultural or scientific topics — local as well as global — can fall under it. A programmatically generated SEO landing page for a product does not meet this threshold under the current position.

3. Two exceptions provide further relief. Even where the public-interest threshold is met, the duty falls away in two cases: legally authorised law enforcement, and where the AI text has undergone a process of human review or editorial control and a natural or legal person holds editorial responsibility for the publication. The draft Transparency Code of Practice tightens this editorial exception: it requires a documented workflow with named responsible persons, not the mere assertion that a review took place.

4. The machine-readable marking mainly hits providers. Article 50(2) requires providers — that is, the model makers — to mark their systems’ outputs in a machine-readable format and make them detectable as artificially generated. That is not the duty of the deployer publishing an SEO page. For generative systems placed on the market before 2 August 2026, this marking obligation was deferred to 2 December 2026 as part of the Digital Omnibus agreement of 7 May 2026.

Interpretation

For programmatic SEO, the situation is therefore clearer than the noise suggests: an AI-generated marketing or product page needs no mandatory label under Article 50(4) because it does not reach the public-interest threshold. This is not a marginal reading but follows directly from the wording and the draft guidelines. Voluntary transparency is unaffected — and can make sense as a trust and GEO argument — but it is not a legal requirement.

More interesting is the borderline case that boostN’s own news module represents. News texts about AI, regulation or market developments touch economic and at times societal matters — they could indeed trigger the public-interest threshold under Article 50(4). Here, though, the second exception applies: if an editorial process with named responsibility is documented, the disclosure duty falls away. That shifts the requirement away from an AI label and towards a verifiable editorial workflow — which is the better editorial standard anyway.

One caveat belongs to the cautious view: both the guidelines and the Code of Practice existed only in draft form in mid-2026. The consultation ran until 3 June 2026, and the final versions were outstanding at the time of writing. The line described here follows the draft stage and should be checked against the final publication.

What you can do now

Treat AI marketing and SEO texts as not subject to labelling — but document the assessment. Record why a page does not reach the public-interest threshold. In case of doubt, that protects better than a precautionary but unfounded label.

Establish a documented review workflow for editorial AI content. Anyone producing news, specialist articles or public-interest content with AI support should actively use the editorial exception: named responsible persons, a traceable review step, recorded in writing.

Check the final version of the guidelines once it is available. The draft stage is robust but not definitive. Plan a re-check of the public-interest definition and the editorial exception.

Entdecke mehr

Blog

Saving Tokens with Claude: 6 Principles That Make Experts Twice as Fast

How I turned my CLAUDE.md from a style guide into a token budget — 6 principles for lower cost, less waiting, and more honest reporting.

Glossar

AI Marketing

AI marketing is the use of artificial intelligence and machine learning to analyse marketing data, automate content and delivery, and optimise campaigns in real time. Its main areas are content, personalisation, analytics and generative engine optimization.

Lexikon

Content Creation with AI: a Clean Workflow

Just generating isn’t enough. The reliable flow: briefing, draft, fact-check, voice, SEO, human final edit — and where the human stays mandatory.

Themenuebersicht

Arbeiten mit KIContent-Erstellung