AI-generated content: when you must disclose it, and how

Since 2 August 2026, the EU AI Act requires whoever publishes AI-generated content to disclose it when that content could be mistaken for something real. If you publish on property portals or social media, it applies to you. This page explains what to do, in five minutes and without jargon.

1. Who has to disclose

The duty to disclose falls on whoever uses the AI tool to create the content and distributes it. The Regulation calls this the “deployer”, and when you publish your listings, that is you. VibePeak, as the provider of the tool, has a different duty: to mark what it generates in a machine-readable way so it can be identified as AI-made. Section 7 explains what we do and what is left to you.

2. Four steps to check whether you must disclose

  1. Was the content generated or modified by artificial intelligence? If you created it with VibePeak, yes. Video, image and audio all count: virtual home staging, a photo with a replaced sky, a video built from photos or a synthetic voice-over are all in scope.
  2. Are you going to publish or distribute it? Posting it on a portal, on social media, on your website, or sending it to a client all count as distribution. Internal use that never leaves your organisation does not.
  3. Does it depict real people, places, objects or events? A specific flat that is on the market is a real place. An identifiable person is a real person.
  4. Could it appear authentic to whoever sees it? If someone could take it for an unedited photo or video, the answer is yes.

If you answered yes to all four, you must disclose.

3. A worked example

You list a real flat for sale on a property portal, using photos to which you applied virtual home staging.

1. AI-generated? Yes. 2. Published? Yes. 3. Real place? Yes — that flat exists and is for sale. 4. Does it look authentic? Yes — realism is precisely what makes home staging useful.

→ You must disclose.

This is the most common case in the sector, and the answer is yes. Acting in good faith and with no intention to mislead does not change the obligation: the rule does not look at intent, it looks at whether the content could appear authentic.

When the answer is no:

  • An obviously unreal image that nobody would mistake for a photograph.
  • Ordinary retouching — exposure, cropping, straightening, white balance — without adding or removing elements.
  • Material that stays inside your company and is never published.

4. If people appear: avatars, cloned voices and AI-added people

An avatar, a cloned voice or an image that reproduces a real, identifiable person (you, a colleague, a client) meets steps 3 and 4 by definition: it always needs the disclosure. On top of that, before publishing you need that person’s consent. That is required not by the AI Act but by data-protection and image rights.

People that AI adds to a room are not real people, but the room is real: that photo needs the disclosure too, because of the place.

5. What counts as a valid disclosure

A valid disclosure meets four conditions:

  • Clear, in language anyone understands: “AI-generated image”, “Video generated with artificial intelligence”, “Virtual home staging”.
  • Visible from the start, not at the end and not hidden behind a “show more”. In a video, a visible mark as playback begins is enough; in an image, text on the image itself or in the caption.
  • Attached to the content, not buried in a general legal notice on your website.
  • Accessible, perceivable by everyone, including people with disabilities: legible text with good contrast, and if you publish alt text or subtitles, mention there too that the content is AI-made. The Regulation requires the disclosure to meet accessibility requirements.

It does not need to be loud or take up half the screen. It does need to be seen by the first person who sees the content.

6. How to do it on each channel

Property portals (Idealista, Fotocasa and similar)

Mark the affected images with text on the image itself and state it in the listing description as well (“The living room and bedroom images include virtual home staging”). If you also publish the property video, the disclosure has to be in the video, because it gets shared on its own.

Instagram, TikTok and Facebook

Use each platform’s AI-generated content label and, on top of that, a visible mark in the video or image itself. The platform label disappears as soon as someone downloads and forwards the file; the mark on the image does not.

WhatsApp and sending files directly to clients

There is no platform label, so the disclosure has to be inside the content or in the message that carries it.

Your own website

A caption or a mark on the image. For video, a visible mark at the start.

Article 50 of the EU AI Act (Regulation (EU) 2024/1689, in force since 1 August 2024) applies from 2 August 2026. It requires whoever deploys the AI, which in your case is whoever publishes the listing, to disclose when an AI-generated or AI-manipulated image, audio or video resembles real people, places, objects or events and could appear authentic. That is the same four-step check as in this guide. Portals can ask for more: Idealista’s help centre asks for photos generated or substantially modified with AI, including virtual home staging, to be clearly and visibly labelled. Check each portal’s rules before you publish.

Why we insist on the visible mark

The photos you edit with AI and download from the app, and the videos of your properties, carry technical provenance information (C2PA content credentials) that identifies them as AI-made. That information is robust, but portals and social networks re-process the files you upload and frequently strip it. The disclosure a human sees is the one that survives that re-processing.

7. What VibePeak does for you

  • Technical marking, automatic. The photos you edit with AI and download from the app, and the videos of your properties, carry C2PA content credentials, the standard Idealista already reads to detect AI-made content. There is nothing to switch on.
  • The visible notice, with one switch. When you create the video, or when you apply elements in the photo editor, turn on “Show the AI notice”. We burn it into the file itself: “AI-generated content” on videos, from the first frame, and “Digitally altered image” on photos, along the top edge. So it survives even if the portal, the social network or WhatsApp strips the metadata. Turn it on when you create: it is a setting of each video and each photo.
  • Your originals, always at hand. Download the original photo alongside the AI version to publish the before and after, or to request a review if a portal flags a photo by mistake.

Besides turning the notice on, the only thing left to you is to mark AI-made photos on the portal when you upload them (on Idealista, when uploading or editing them) and to use each social network’s AI label. By law that decision belongs to whoever publishes, and no tool can make it for you.

8. One caveat

This page is informational and does not constitute legal advice. It covers the ordinary case of a real estate agency publishing listings in the European Union. If your case falls outside that (advertising outside the EU, regulated sectors, or doubts about the consent of a person who appears in the content), consult your own legal advisers.

Last updated: 15 October 2026.