IMScollect
On a cream-coloured work surface lie an unlabelled checklist, a smartphone with a blank screen and several media cards with orange check marks.
Checklist

AI Act checklist: labelling chatbots, AI text and images

Interactive AI Act checklist for SMEs: roles, labelling, human review and evidence for chatbots, AI-generated text, images and deepfakes.

ISO 9001ISO 14001ISO 27001Legal requirements
Philipp Kappe

Philipp Kappe

Founder of IMScollect

Published 27 August 2026
Contents

Chatbots, AI-generated text, images and videos are not all subject to the same labelling obligation across the board. What matters is the specific use case, the company's role and whether a human actually reviews the output and takes editorial responsibility for it. This article puts the assessment into context; the complete interactive checklist then guides you through decision, evidence and action planning.

What has applied since August 2026

The transparency obligations under Article 50 of the European AI Regulation (AI Act) have applied since 2 August 2026. They distinguish between obligations of the providers of an AI system and obligations of the companies that use such a system.

Use caseKey questionPossible obligation
Chatbot or voicebotIs it obvious to a reasonably well-informed and observant user that they are interacting with AI?Providers must generally enable the information to be given at the latest at the first interaction.
AI-generated image, audio, video or textIs the company the provider of the generating system?Synthetic outputs must generally be markable in a machine-readable format.
DeepfakeCould the content falsely appear real or authentic?Deployers must disclose that the content has been artificially generated or manipulated.
AI text on matters of public interestWas the text published with effective human review and editorial responsibility?Without these conditions, disclosure may be required.

Not every AI output needs the same visible notice. Machine-readable marking on the provider side is something different from visible disclosure by a deployer. A documented case-by-case decision is therefore more robust than the blanket statement “created with AI”.

Provider or deployer: the role is determined for each application

A provider is anyone who develops an AI system or has it developed and places it on the market or puts it into service under their own name or trademark. A deployer is anyone who uses an AI system under their own authority. Depending on the application, a company can have one or both roles.

With a purchased website chatbot, the technical information obligation under Article 50(1) generally lies with the provider of the directly interacting system. The company using it should nevertheless check whether the notice is visible in the actual context of use and how technical and contractual responsibilities are allocated.

The role should therefore not be set once across the board for the whole company. A third-party chatbot, an internally developed application and an AI-supported editorial process may need to be classified differently.

Six questions for the initial check

These questions show whether a more in-depth assessment is required:

  1. Do customers, applicants or other people interact directly with the AI system?
  2. Can the person concerned tell, without an additional notice, that AI is involved?
  3. Does the system generate or modify text, images, audio or video?
  4. Could content falsely appear real or authentic?
  5. Is an AI text published to inform the public on matters of public interest?
  6. Is there an actual human review process and a clearly named person with editorial responsibility?

A robust decision also needs information on the system provider, the technical marking, the publication channel and possible exceptions. The complete checklist brings exactly this information together.

What the complete PDF contains

The complete working document includes:

  • recording of the AI use case and the company's role
  • ten check questions on labelling, technical marking and deepfakes
  • separate review of AI texts and editorial control
  • evidence register with storage location, responsibility and next review date
  • concise classification for ISO 9001:2026, ISO 14001:2026 and ISO 27001:2022
  • overall decision and action plan
  • 158 interactive form fields, including genuine option groups for “Yes”, “No” and “Open”

The fields support umlauts and other special characters, can be saved and work without PDF JavaScript. The checklist can therefore be completed digitally or printed out and used in a review.

From individual case to integrated compliance process

The AI Regulation does not come from an ISO standard. Even so, a company does not need an isolated special process for it. Applicability, obligation, responsibility, action, evidence and next assessment can be managed in one shared legal and compliance process.

Management systemUsable contributionSubject-matter limit
ISO 9001:2026Process ownership, competence, documented information and change controlThe AI Act does not automatically become a quality requirement. What matters is the link to products, services and customer processes.
ISO 14001:2026Proven system for compliance obligations and evaluation of complianceThe AI Act is generally not environmental legislation. Environmental compliance obligations remain separately identified.
ISO 27001:2022System inventory, legal requirements, suppliers, changes and protection of evidenceAn ISMS structures the process but does not replace the legal review of roles and obligations.

Integration means one shared management process. It does not mean artificially assigning every piece of legislation to every standard. The white paper “Integrated management system: what ISO 9001, 14001 and 27001 really share” shows where shared processes work and where the subject-matter separation must remain.

Conclusion

The transparency obligations cannot be dealt with by a general AI notice. First, the use case and the role must be clarified. This is followed by the labelling decision, technical or visible implementation, human approval and robust evidence.

The interactive checklist brings these steps together in a fillable working document. IMScollect can then link the resulting responsibilities, actions, documents and review dates. The legal assessment of the specific use remains the responsibility of the company and its professional advisers.

Frequently asked questions

Do all AI-generated texts have to be labelled?▾

No. For deployers, Article 50 covers in particular published AI texts intended to inform the public on matters of public interest. Disclosure may not be required if genuine human review or editorial control has taken place and a person holds editorial responsibility.

Does a chatbot have to say that it is an AI?▾

Article 50(1) requires the provider to design directly interacting AI systems so that people are informed that they are interacting with AI, unless this is obvious anyway. The notice must be clear, accessible and given at the latest at the first interaction.

When does an AI image or deepfake have to be labelled?▾

Providers of generative systems must make certain synthetic outputs technically markable. Deployers must generally disclose deepfakes; the specific form and possible special rules for artistic or fictional works must be checked against the use case.

What is the difference between provider and deployer under the AI Act?▾

Providers develop an AI system or place it on the market under their own name. Deployers use an AI system under their own authority. Depending on the use case, a company can have different roles.

What evidence should an SME keep on AI labelling?▾

Useful items are an AI inventory, the role assessment, the labelling decision with its reasoning, the wording of the notice used, technical documentation and evidence of human review, editorial responsibility and approval.

Sources

Checklist for download (in German)

PDF, 6 pages

The text of this article is in English. The document itself is only available in German so far; you can request it free of charge on the German page of this article.

Download on the German page (in German)
Philipp Kappe

Philipp Kappe

Founder of IMScollect

Philipp builds IMScollect. TÜV-certified auditor with a background in governance, risk and compliance, more than 200 audits conducted and supported.

By email