Your AI twin just delivered a flawless keynote recap on LinkedIn. The lighting is perfect, the messaging is sharp, and the engagement is climbing. There is just one problem: nowhere in that post does it say the voice cloning technology behind it exists.
That gap is no longer a minor oversight. With the EU AI Act's Article 50 transparency obligations taking effect in August 2026 and UK regulatory expectations tightening around synthetic media, professionals using AI video twins for personal branding are entering a compliance landscape that has real teeth. Crucially, the rules do not hinge on whether you intended to deceive anyone. If your content realistically depicts you using AI generation or manipulation, disclosure is required, full stop.
This guide cuts through the legal noise and gives you exactly what you need to act today. You will learn what qualifies as a twin video under current definitions, what a valid disclosure actually looks like in practice, the precise caption language that meets compliance standards, and how to build these requirements into your content workflow without slowing your output. Confidence and compliance are not in conflict. This post shows you why.
Why Disclosure Rules Matter Right Now
On 2 August 2026, EU AI Act Article 50 transparency obligations became legally enforceable. Disclosing AI-generated content is now a legal requirement, not a recommended practice.
The European Commission published its final guidelines on 20 July 2026, defining "deepfake" broadly as any content that falsely appears authentic or truthful. That definition includes realistic AI-generated depictions of real people, so a professional twin video falls within scope.
UK professionals are not outside this framework. If you post content that reaches EU audiences, Article 50 applies to you as the deployer. Using an AI system in a professional context without substantially modifying it makes you the deployer under the regulation.
The UK's own regulatory direction is also tightening. The FCA's principles-based approach means consumer-facing professionals are expected to handle AI disclosure under existing consumer protection and transparency frameworks, even without dedicated synthetic-media rules.
If you want to understand the technology behind these obligations before working through compliance steps, this guide to how AI voice cloning works in a personal branding tool covers the basics plainly and practically.
What Counts as a Twin Video Under the Rules
So what, precisely, triggers these obligations for your content?
A twin video is a realistic AI-generated depiction of your own face and voice. The technology produces video of you speaking a specific script without you filming that clip. That output meets the European Commission's definition of a deepfake. Consent and accuracy do not change this.
Artificial intelligence voice cloning carries the same obligation, even without a visual twin. If your cloned voice delivers lines generated by a model trained on your recorded voice, that audio is synthetic and must be disclosed. This applies whether the voice appears in a video, a standalone audio post, or a podcast clip. For a detailed look at how voice cloning laws treat this distinction, the article on AI voice cloning laws and why cloning your own voice is the safe move is worth reading before you publish.
Mixed formats are not exempt. A post combining a real photograph with an AI-generated voice-over triggers the same rules as a full twin video. Any single component generated by an AI system that produces a realistic depiction of a real person is enough.
The rule is simple: if AI produced any part of how you look or sound in a post, disclosure is required.
What Does Not Count as a Valid Disclosure
Knowing what triggers the obligation is only half the picture. Many professionals assume they are covered when they are not.
Metadata alone fails. Embedding an AI flag in a file's properties is invisible to any viewer watching on LinkedIn or Instagram. Disclosure must be visible at the point of interaction, not stored in technical attributes no one will ever open.
Unreadable labels fail. A label rendered in low contrast against a busy background, or positioned where an ordinary viewer on a mobile screen cannot read it without effort, does not meet the Commission's clearly visible standard.
Profile bios and old pinned posts fail. A line in your bio saying you use AI tools, or a disclosure post from three months ago, is not attached to the specific content being viewed. The obligation is per post, not per account. Referencing AI-generated content and disclosure in your terms and conditions carries the same problem: it is not visible at the moment a viewer watches your twin video.
Vague language fails. Phrases such as "created with AI tools" or "AI-assisted content" do not tell the viewer that the face or voice they are watching is AI-generated. The notice must say that clearly and directly.
A quick colleague review does not unlock the editorial exception. Article 50(4) provides a narrow exemption requiring genuine, substantive editorial oversight with documented accountability. An informal check before posting does not meet that standard.
The Three Elements Every Compliant Post Needs
Now that you know what will not pass scrutiny, here is what a compliant post must contain.
1. A plain-language written notice in the caption or on-screen text. State clearly that the video is AI-generated. The wording must be unambiguous and must appear at the point the viewer encounters the post. "Made with AI tools" does not meet this standard. "This video features an AI-generated likeness and voice" does.
2. A persistent visual label in the video itself. A persistent visual label in the video itself, such as "AI-generated video" or "AI twin" at the top or bottom of the frame, is required. The Commission's plain-language, clearly visible standard implies the label must be genuinely readable, not a brief flash that a viewer is likely to miss.
3. An audio cue where artificial intelligence voice cloning is used. When the voice is AI-generated, the viewer must be told this in a way they can hear or clearly read before or during playback. A brief spoken or on-screen introduction at the start of the video covers this.
All three elements must appear together. A caption notice without an on-screen label, or an on-screen label without a caption notice, leaves a gap that regulators could find insufficient.
The notice must also sit within the post itself. A comment added after publishing, or a Story that expires, does not count. On LinkedIn and Instagram, disclosure belongs in the post at the moment the viewer sees it.
Compliant Caption Language You Can Use Today
Knowing the three required elements is one thing. Knowing exactly what to type is another. Here are the phrases you can copy, paste and adapt today.
For a full twin video, where both the face and voice are AI-generated, use this as the opening of your caption:
"This video features an AI-generated likeness and voice. The views, experience and expertise expressed are entirely my own."
For a post using artificial intelligence voice cloning only, where the visual is real but the voice is synthetic:
"The voice in this video is AI-generated using voice cloning. The content reflects my own views and knowledge."
As a practical matter, most mobile feeds collapse long captions. Placing your disclosure in the first two lines of your caption is the safest approach to ensure the notice is visible at the point of interaction, as the regulation requires.
Apply the on-screen label described in the previous section.
The opening audio cue described earlier applies here too.
Platform Settings That Support Your Disclosure
Caption language gives you the words. Platform settings give you a second layer on top.
LinkedIn includes an AI content label in its post creation flow. Select it for every twin video post. It sits alongside your caption notice, it does not replace it.
Instagram offers a "Made with AI" label in the advanced settings of a post. Apply it each time. Treat it as supplementary, because the platform controls whether that label persists, changes format, or disappears in a policy update. Your caption wording stays constant regardless.
Facebook provides similar labelling options in the post composer. Apply the label, and include your written disclosure in the caption using the same plain-language wording covered in the previous section.
The critical point is this: none of these built-in labels currently guarantees compliance with Article 50's plain-language, point-of-interaction standard. Regulators have confirmed that technical labels alone, without a clearly visible written notice the viewer can read at the moment they see the post, are not sufficient. Your caption is the element you control fully. The platform label is a welcome addition, not a safety net.
Check each platform's AI labelling policy before you publish. These settings change without notice, and the threshold for what triggers mandatory labelling is still evolving alongside regulatory guidance. For Brandios's own platform-specific commitments, review the privacy page before connecting your accounts.
The Editorial Exception and Whether It Applies to You
Some professionals ask whether the editorial exception in Article 50(4) removes the need to disclose. It does not, for most of you.
The exception covers AI-generated text used to inform the public on matters of public interest, where a named individual assumes genuine editorial responsibility. A quick review before posting does not meet that standard. The guidelines require substantive oversight with clear accountability, a bar designed for institutional journalism, not a solo professional approving their own content.
More importantly, the exception does not apply to AI-generated video. Even if your post is genuinely educational, the requirement to disclose a realistic AI depiction of your face or voice stands. The visual and audio deepfake disclosure obligation is a separate rule, and no editorial exception touches it.
A post promoting your coaching practice, sharing a market view, or building your personal brand does not qualify as public-interest publishing under any reasonable reading of the exception.
Do not rely on Article 50(4) without taking specific legal advice. The compliance threshold is higher than most professional content workflows currently meet, and the safer position is to disclose every time.
Building Disclosure Into Your Content Workflow
The question becomes practical: how do you make compliant disclosure automatic rather than something you remember on a good day?
Start with a saved template. Write your standard disclosure phrase once, save it in your notes app or a document you open easily, and paste it into every caption that accompanies a twin video. One phrase, used consistently, removes the risk of forgetting.
Own the responsibility at approval. If you use a personal branding tool that prepares twin videos and captions for you, the compliance obligation sits with you as the deployer, not the platform. Read the caption before you approve any post. Confirm the disclosure is there.
Brandios builds twin videos from a single clip you record once, using your face and AI voice cloning, then sends each video to you for approval before anything publishes. That approval step is your compliance checkpoint. Use it deliberately.
Apply a three-point check before every approval:
Caption includes a plain-language AI notice in the first two lines
Video file carries a persistent on-screen label
Platform AI label setting is applied in the post composer
Review quarterly. Regulatory guidance is still developing; set a quarterly reminder to check whether the disclosure language you are using remains consistent with any updated Commission or FCA guidance. Rapid AI adoption is why regulators are watching deployer behaviour closely, and standards can shift between review cycles.
A Compliant Post Is a Confident Post
Once your workflow is in place, the only question left is confidence.
Transparent disclosure does not weaken your professional standing. It demonstrates it. Saying clearly that a video features your AI-generated likeness and voice shows your audience you are honest, informed, and in control of your content. That is exactly the impression a credible professional wants to make.
Done consistently, the steps covered in this guide keep you on the right side of the rules and, more importantly, on the right side of your audience.
If your goal is a consistent, reliable presence without filming every day, the route there is openness. Audiences who trust that you will tell them what they are watching will stay with you as AI content becomes more common, not less.
Take the caption template from the earlier section, add it to your workflow today, and revisit it as UK-specific rules continue to develop. The regulatory picture will sharpen. Your disclosures should sharpen with it.
If you are ready to build that presence properly, Meet your twin.
Conclusion
The rules are still developing, and professionals who build disclosure into their workflow now will adapt easily, while those who treat it as optional will face increasing risk.
Your audience is already watching AI content. What they remember is who was honest with them.
