Why AI video labeling matters now
Generative video is becoming part of ordinary marketing production: a team may use an AI tool to create a product scene, alter footage, generate an avatar, or produce a complete advertisement. On August 2, 2026, new transparency obligations under Article 50 of the European Union’s AI Act become applicable. The European Commission says the rules are intended to help people recognise when content has been generated or altered by AI and to reduce deception and manipulation.
This is not a general ban on AI video. It is a transparency framework. The practical question for a marketing team is whether a particular video is AI-generated or manipulated content covered by the disclosure rules, and, if so, how to make that disclosure clear and detectable.
The Commission’s official announcement says deepfakes—AI-generated or edited images, videos, or audio—must be labelled. It also says AI-generated or altered content must carry machine-readable marks so it can be detected more easily. The precise obligation depends on the role a business plays, the kind of content it publishes, and the circumstances of publication.
> This article is an operational overview, not legal advice. Teams serving EU audiences should review the official guidance and obtain qualified advice for their specific campaigns.
What the EU AI Act says about AI-generated video
Article 50 separates several transparency situations. For video marketers, the most important is the deployer obligation concerning deepfakes. The Commission defines a deepfake for this purpose as AI-generated or manipulated image, audio, or video content that resembles existing persons, objects, places, entities, or events and would falsely appear authentic or truthful to a person.
That definition is narrower than “any video that used AI.” For example, an obviously fictional animation may not present the same authenticity risk as a synthetic video that appears to show a real person giving a real testimonial. However, a team should not assume that a particular style, filter, or disclosure elsewhere in a campaign automatically resolves its obligations.
The Commission’s guidance also addresses providers and deployers:
- A provider develops an AI system, or has one developed, and places it on the EU market or puts it into service under its own name or trademark.
- A deployer uses an AI system under its authority for professional or other non-personal activity. An advertising company using a system for a client can therefore have responsibilities as a deployer.
- Providers outside the EU can still be subject to relevant AI Act provisions when the output of their AI system is used in the EU.
A business commissioning an AI advertisement should map its role and responsibilities rather than assuming that the software vendor handles every disclosure requirement.
Which marketing videos should teams review first?
Begin with videos where viewers could reasonably mistake synthetic content for authentic footage. Prioritise the following inventory:
- Synthetic people or spokespersons. This includes an AI-generated person presented as a customer, employee, expert, or public figure, particularly when the video resembles a real testimonial or interview.
- Altered real people. Face replacement, voice cloning, lip synchronisation, or edits that make a real person appear to say or do something they did not say or do deserve immediate review.
- Fabricated places, events, or product demonstrations. A generated scene that looks like a real location, live event, news moment, or factual product experience can create an authenticity issue.
- AI-generated audio paired with realistic video. A synthetic voice can change how a viewer interprets an otherwise ordinary clip, especially when it imitates a named individual.
- Repurposed campaign variations. A master video may be labelled while an automatically generated cut-down, translation, or new version is not. Treat each delivered asset as part of the same audit trail.
The Commission’s icon guidance says not all AI-generated or manipulated content needs to be labelled under Article 50(4). It specifically identifies deepfakes and certain AI-generated or manipulated text published to inform the public on matters of public interest. That distinction is why a documented review is more useful than a blanket claim that every AI-assisted edit has identical treatment.
A practical labeling workflow for AI video ads
A repeatable workflow can make compliance part of production instead of a last-minute blocker.
1. Record how the asset was made
For every video, capture the tools used, the kind of transformation performed, the source footage, and the person responsible for approval. Note whether AI generated new pixels, changed a person’s appearance or voice, created a setting, or only assisted with tasks such as trimming or caption drafting.
2. Test for an authenticity risk
Ask a simple review question: Could an ordinary viewer believe that this video shows a real person, place, event, or statement? If yes, escalate it for a disclosure decision. Reviewers should consider the video, audio, caption, thumbnail, landing page, and surrounding ad copy together.
3. Choose a visible disclosure
When disclosure is required, make it clear and distinguishable. The placement should not be hidden in an unrelated privacy page or made unreadable by the video’s background. For a short-form ad, a label at the beginning, a persistent on-screen notice, or another appropriate presentation may be more useful than a disclosure that appears only after the viewer has finished watching.
The Commission has published freely available EU icons that creators, publishers, and other deployers may use for certain AI-generated content. The icons are available in black, white, and transparency variations, with downloads in SVG and PNG formats. They are a practical option, but teams should still check whether the context calls for additional plain-language explanation.
4. Preserve machine-readable marking where applicable
The Commission’s announcement says AI-generated or altered content must also carry machine-readable marks so it can be detected more easily. Do not remove provenance or metadata during editing, resizing, transcoding, or platform export without understanding the effect. Keep a high-quality master and record what happens to markings at each delivery stage.
5. Approve, archive, and monitor
Store the final labelled file, the uncompressed master, the tool record, the approval decision, and the destination markets. If a platform strips metadata or changes the creative, document the version that was actually published. Re-check the asset when it is translated, materially edited, or used for a new audience.
How to use the EU Code of Practice
The European Commission describes its Code of Practice on Transparency of AI-generated Content as a tool that supports compliance with Article 50 obligations. It has two sections: one for providers, covering marking and detection of AI-generated and manipulated content, and one for deployers, covering labelling of deepfakes and certain AI-generated or manipulated text.
The code itself is voluntary, while the Article 50 transparency requirements are legal obligations. The Commission and the AI Board have confirmed the code as an adequate voluntary tool for demonstrating compliance. Providers and deployers that sign it can rely on its measures for the relevant labeling and detection requirements; organisations using other means need to be able to demonstrate that their approach is adequate.
For a marketing team, the useful lesson is straightforward: adopt a documented method, align it with the official guidance, and be able to explain why each asset was or was not labelled. A checklist, asset register, and approval owner are more valuable than an informal promise that an editor will remember the rule.
What this changes for AI video performance marketing
Transparency does not replace creative strategy. It changes the production brief. Teams should plan disclosure space before editing, test whether labels remain legible on mobile placements, and avoid creative concepts that depend on viewers believing a synthetic scene is genuine.
A sensible campaign process can include:
- A disclosure field in the creative brief and asset naming convention.
- A human review for synthetic people, voices, testimonials, and realistic events.
- Separate masters for different aspect ratios and placements.
- Platform-by-platform checks for visible labels and metadata preservation.
- A record of countries, audiences, landing pages, and publication dates.
- A post-launch check to confirm that the published version matches the approved version.
These steps also improve trust. A clear disclosure gives viewers context and helps a brand avoid an avoidable surprise when an audience discovers that a supposedly authentic moment was generated or manipulated.
FAQ: AI video labeling and the EU AI Act
Does every AI-assisted marketing video need an EU AI Act label?
No blanket rule should be inferred. The Commission’s icon guidance says the disclosure requirement covers specific AI-generated or manipulated content, including deepfakes, and certain AI-generated text on matters of public interest. Review the nature of the transformation and whether the result could falsely appear authentic or truthful.
When do the Article 50 transparency obligations apply?
The European Commission states that the Article 50 transparency obligations apply from August 2, 2026. This article uses that date because it is the date in the Commission’s current guidance and announcement.
Can a brand use the EU AI icons on a video ad?
The Commission says creators, publishers, and other deployers may use the freely available icon set to label certain AI-generated content. Teams should use an appropriate, clear disclosure and confirm that the final placement remains understandable in the relevant format.
What if an AI tool or agency created the video for us?
Identify who is the provider and who is the deployer under the facts of the project. The Commission’s FAQ explains that a legal person using an AI system under its authority can remain the deployer even when contractors or freelancers help operate it. Allocate review, labeling, and record-keeping responsibilities in writing.
Is the Code of Practice mandatory?
The Code of Practice is voluntary, but the transparency requirements under Article 50 are legal obligations. The code is one way to support and demonstrate compliance; an organisation choosing another method should be ready to show that its measures are adequate.
Zapplon helps businesses turn AI into practical growth systems through AI video production, AI agents, and performance marketing services. We can help you plan, create, review, and deploy campaign assets with a clear workflow. Services start at $50. Contact Zapplon to discuss your next campaign.