Adapte Dijital
Kurumsal
Müşteri Girişi
Dijital Yönetim
AI SEO
Marka Yönetimi
Danışmanlıklar
Web & App & AI
Ads & Reklam
Kitle Yönetimi
Veri Yönetimi
Amaç & Hedef
Videolar Referanslar Dijitale Yatırım Yap
Blog
Machine Trust and Compliance

What Is the AI Content Code of Practice? The EU’s Labelling Guide in Plain Language

AuthorTeknoloji Masası Published21 November 2026 Reading Time6–10 dk
What Is the AI Content Code of Practice? The EU's Labelling Guide in Plain Language
💡 Kısaca: The AI content code of practice is the voluntary guide the European Commission published for marking and labelling AI-generated content.

The AI content code of practice is the voluntary guide the European Commission published for marking and labelling AI-generated content. It explains step by step how the transparency rules of Article 50 of the AI Act are applied in practice. 📘

This guide explains what the code is, what it asks of providers and of businesses that use AI, what it means in practice for a Turkish business and which misunderstandings are common. The information was compiled from the Commission’s announcement and legal sources on 12 November 2026. This article is for information only and is not legal advice; consult a lawyer about your own situation.

In short: The core of the code: providers apply layered marking and offer detection tools, deploying businesses label with common icons. The Act itself is summarised in our EU AI Act Article 50 summary.
WHAT

What is the code of practice?

BU BÖLÜMÜN ÖZETİ

  • A voluntary roadmap
  • The adequacy assessment
  • Its relation to the law

The code turns the general wording of the law into workable steps. Knowing what it is, and what it is not, is the precondition for using it well. The wider picture of labels, provenance and logs is set out in the machine trust and compliance guide.

Status of the EU AI content code of practice
The code is a tool, not an exemption.

A voluntary roadmap

The Commission published the final code on 10 June 2026, after a first draft in December 2025 and a second in March 2026. It was drawn up by independent experts through a process that, according to the Commission, involved more than 187 participants. Adopting the code is voluntary, but the obligations of the Act bind everyone, whether or not they sign. 🗺️

The code turns the general wording of the law into workable steps.

The adequacy assessment

According to legal sources, the Commission on 8 July and the AI Board on 9 July 2026 concluded that the code is adequate to facilitate the practical implementation of Article 50’s marking, labelling and related provisions. That makes the code a recognised way to demonstrate compliance. After the general-purpose AI code, it is the second code to receive a positive assessment.

Its relation to the law

The code does not replace the Act, nor the Commission’s Article 50 guidelines; it complements them. Signing it does not change other obligations, such as those for high-risk systems or general-purpose models. The code is a tool, not an exemption. Not signing does not release anyone from the obligations either. ⚖️

WHAT

What does it ask of providers?

BU BÖLÜMÜN ÖZETİ

  • Layered marking
  • Detection tools
  • No single technical solution

The first section of the code is addressed to providers of AI systems that generate content.

Layered marking

According to legal sources, the final code requires at least two layers of machine-readable marking to meet Article 50’s marking duty, for example signed metadata combined with an invisible watermark. More layers reduce the risk of a marker being lost on the way. If one layer is stripped, the other may survive, which is the whole point of layering. 🧱

The first section of the code is addressed to providers of AI systems that generate content.

Detection tools

The code asks providers to keep available a solution for detecting content generated or manipulated by their systems. Platforms, journalists and users can then check where a piece of content came from. A detection tool gives the marker its meaning; a marker that cannot be read produces no trust, however well it is embedded.

No single technical solution

According to commentators, no marking technique today meets all four of the Act’s requirements at once: effectiveness, interoperability, robustness and reliability; that is why the code does not impose a single solution. The layered approach is the practical answer to this technical limit. How the main provenance standard works is explained in what C2PA content credentials are. 🔧

WHAT

What does it ask of deploying businesses?

BU BÖLÜMÜN ÖZETİ

  • Common EU icons
  • Placement and design of the icon
  • Publishing text

The second section of the code is addressed to businesses that use AI in their own work and publish content.

What the code asks of providers
More layers reduce the risk of loss.

Common EU icons

The Commission published free common icons for labelling AI-generated content. The icons let readers understand the nature of the content even without knowing the language. A common icon reduces confusion between labels. Instead of every business inventing its own label, a recognised sign is used, which also makes the label easier for platforms to read. 🏷️

The second section of the code is addressed to businesses that use AI in their own work and publish content.

Placement and design of the icon

According to legal sources, the final code dropped the first draft’s common taxonomy and focuses instead on detailed requirements for the design and placement of icons, labels and disclaimers. The label should be visible, legible and close to the content. A label squeezed into a corner, or too small to read, defeats its purpose. How to build a labelling routine is covered in setting up an AI content labelling system.

Publishing text

AI-generated text published to inform the public needs a label, but the duty does not apply if the text has undergone human review and someone holds editorial responsibility. The editorial review should be documented. If the record shows who reviewed and approved it, the exception can be defended when a question comes. 📝

WHAT

What it means in practice for a Turkish business

The code is a European document, but it offers a Turkish business three practical lessons.

Choosing tools

Because the code asks providers for layered marking and detection tools, whether a business’s AI tools meet these conditions becomes a selection criterion. A tool that adds no marker may cause trouble later for content that reaches Europe. Asking the provider in writing is the cleanest way to find out, and internal AI use reviews help, as described in business AI usage audit.

The code is a European document, but it offers a Turkish business three practical lessons.

A labelling routine

The code’s common icons and placement rules can serve as a model for a business’s own labelling routine. The same logic can be used for the disclosure provision of Türkiye’s advertising regulation. One routine for both rules is simpler than two. The routine is written into the AI usage policy and applied by the person who owns machine trust. 🛠️

COMMON

Common misunderstandings

BU BÖLÜMÜN ÖZETİ

  • The code makes you label everything
  • No signature, no obligation
  • Set it up once and you are done

Three misunderstandings about the code come up often. All three lead to the wrong preparation.

What the code asks of deploying businesses
Reading the code matters more than signing.

The code makes you label everything

No. Article 50 imposes duties on particular kinds of content and particular actors; most routine AI-assisted work falls outside them. The Commission’s final guidelines also broadened the exceptions to the marking duty. Labelling without knowing the scope makes labels meaningless. 🚫

Three misunderstandings about the code come up often.

No signature, no obligation

No. The code is voluntary, but Article 50’s obligations bind regardless of whether a business signs. The code is one route to compliance; it is not compliance itself. A business that does not sign still has to show its compliance in other ways, which is why a written record of decisions matters.

Set it up once and you are done

No. The code, the guidelines and enforcement practice will evolve, and technical solutions are changing too. Labelling and marking arrangements should be reviewed regularly. A yearly check at least shows whether the rules and tools have changed, and it keeps the policy honest. 🔁

HOW

How does Adapte Digital use the code of practice?

BU BÖLÜMÜN ÖZETİ

  • A single delivery package
  • What is delivered
  • Where to start

We turn the code into a checklist, apply it to your content flow and hand over the labelling routine; legal assessment is done with your lawyer.

A single delivery package

Adapte Digital does not charge a monthly management fee. Work starts with a single delivery package whose scope is written down in advance: 36,000 TRY + VAT. If the work grows, the new need is defined as a separate add-on package, priced by what will be delivered rather than by the hours spent. 📦

We turn the code into a checklist, apply it to your content flow and hand over the labelling routine; legal assessment is done with your lawyer.

What is delivered

The output is labelling rules based on the code, questions for your tool providers and a labelling step added to your content flow. The checklist stays with your team for later reviews.

Where to start

Which packages are open, and at what price, is shown on our campaign pages, item by item. 🧭

Thinking of building a business in this area? Our article on content provenance records looks at that door separately.

QUICK

Quick Summary

  • The code was published on 10 June 2026 and found adequate in July.
  • The code is voluntary; Article 50 obligations are binding.
  • Providers are expected to use at least two marking layers and offer detection tools.
  • Deploying businesses have common EU icons and placement rules.
  • The code does not require labels on everything; this article is not legal advice.
SHORT

Short Glossary

Code of practice
A voluntary guide explaining how a legal obligation is met in practice.
Layered marking
Using more than one marker together, such as signed metadata and an invisible watermark.
Detection tool
A tool that checks whether content was generated by a particular AI system.
FREQUENTLY

Frequently Asked Questions

NEXT

Next Step

Ask the providers of your AI tools in writing how they mark their outputs and whether they offer a detection tool. Our campaign pages list the delivery packages that cover this work.

Sources: European Commission — Code of Practice on marking and labelling AI-generated content · Freshfields — The final Code of Practice on transparency of AI-generated content · Faegre Drinker — Commission confirms transparency Code of Practice as adequate

Updated: November 2026

Adapte Digital · AINEO-assisted desk · Reviewing editor: Mehmet Şenol

Ask the providers of your AI tools in writing how they mark their outputs and whether they offer a detection tool.

Sık Sorulan Sorular

Is signing mandatory?

The signatories are mostly tool providers and large publishers. A small business is not required to sign; but the code’s section for deploying businesses is the clearest guide on how to label. Reading the code matters more than signing it. A one-page summary for the team is often all a small business needs. 📖

Do I have to sign the code of practice?

No, the code is voluntary. But Article 50’s obligations are binding, and the code is a practical way to meet them.

Does the code apply to businesses in Türkiye?

The code is a European document. It guides content that reaches Europe, and it is also a good model for the disclosure provision of Türkiye’s advertising regulation.

Is using the common EU icons mandatory?

The icons were published to make labelling easier. Assess with a lawyer which format suits you.

Bu Konuyla İlgili Diğer İçerikler

Share this article
WhatsAppXLinkedInFacebook

Comments

TREN