“labeling requirement for AI-generated or significantly manipulated images, provided they appear authentic and there is a risk of misleading the viewer.”
In short form it means that an AI image has to be marked as “AI Content” on all websites. The regulation is already in place since 2. August 2026. I don’t want to discuss the merits or flaws of such a regulation here, but this is something we, as website operators in the EU, must take into account. And to prevent that we are getting mail from dubious lawyers we should take that regulation serious.
We all can now of course start to manually mark all our AI pictures and publish them with a batch or so. But it would be good if we have a support within Elements.
What about a possibility to add a checkbox in the resource info tab to mark it as AI content.
And then a possibility in all image related components to publish a batch or so on the image when the checkbox is set.
The intention here is to raise the awareness that such a regulation exists and is already in place (I personally was not aware) and look for solutions.
What exactly requires labelling
The phrasing “appear authentic … risk of misleading” refers to Art. 50(4) of the AI Act (deepfake obligation for deployers) — not every AI image:
• Required: AI-generated or -manipulated images that resemble real people, places, objects, entities, or events and could falsely appear authentic (e.g. photorealistic practice/office photos, “real-looking” team photos, stock-photo replacements from Midjourney/DALL·E)
A checkbox would be nice, we just have to make sure any marking is also WCAG compatible, as this is also a legal requirement. Good would be a marker over the image, which only appears when a checkbox is checked - but this will be a problem for Gallery with a remote folder. In this case we would need a certain word or phrase in the title, so Elements can then add this marker.
First, the good news for everyone outside the EU: this doesn’t apply to you. Grab some popcorn and enjoy the discussion
Same as in my first post: a checkbox in the Resource Info tab, and an option in the image components to output a small badge when it’s set. No image processing, no watermarking pipeline on your side. But in the end I’ll leave the how entirely to you - you know best what’s feasible in Elements.
Yes, since last month. The Commission published its final Article 50 guidelines on 20 July 2026, along with a Code of Practice and an official EU icon set (SVG and PNG, free to use, no attribution required). The placement rules are spelled out too: perceivable at first exposure, not hidden behind overlays, alt/ARIA text where possible. Embedding into the file is preferred, but a UI overlay is explicitly accepted as an equivalent - which is the practical route anyway, since a burned-in watermark is unreadable at mobile breakpoints.
One clarification to my own opening post: it is not a blanket “label every AI image” rule. It targets AI content that resembles real people, objects, places or events and could pass as authentic, with a lighter regime for obviously artistic or fictional work - the test is whether the image could be mistaken for a real photograph, not how important the subject is
As I wrote at the start, my main goal here was awareness - I only came across this by accident myself. I’m not a lawyer, and for the ~25 AI images on my preview pages I can handle it manually, so I’m not really asking for myself. But a fair number of people here build sites for EU clients, and warning letters from law firms will probably arrive long before any regulator does. If the answer is “not an Elements problem”, that’s fine too - I’d just rather know it now than later.
One-time linking — a JavaScript file is loaded via the project’s Resources folder and linked to every page through the Site Template.
Marking individual images — an image that has been AI-generated or AI-edited is given a recognisable CSS class via the Image component’s Inspector.
Marking images inside a gallery — since a gallery has no per-photo setting, marking there happens through a fixed addition to the filename before uploading.
Automatic display — when the page loads, the script recognises the marked images and adds a visible label to each, both visually and in the alternative text for screen readers.
Starting September 27, 2026, a new EU law on environmental claims will affect any property listing that can be viewed or booked by travelers in the EU—even if the property itself is located outside the EU.
Under the new law, environmental and sustainability claims shown to EU travelers will be subject to stricter requirements. If you display an environmental or sustainability claim to EU consumers, it must be substantiated by clear, verifiable evidence or be based on a certification scheme that meets the criteria of the EU Empowering Consumers for the Green Transition Directive (ECGT).
Not a long stretch to imagine this EU ruling on marking of AI images will eventually spread to all sites visible in the EU.
I for one do not want to be conned into purchasing something based on an AI generated image whether it is a product, a holiday, a wife etc etc. False advertising at its very core.
Thanks @CyberFredje, I’ll take a closer look at that. Looks like a workable route with little effort – and it also shows that this doesn’t have to be complicated.
Both would be a nice addition, but as I read the rules neither is enough on its own – happy to be corrected by anyone with deeper knowledge. The disclosure has to be perceivable to the visitor at first sight, and a filename never is. The alt tag actually does appear in the EU guidance, but as an accessibility addition on top of the visible label, not as a replacement for it.
There is a second layer that works roughly the way you describe – machine-readable marking in the file metadata. But that one is the AI provider’s job (OpenAI, Midjourney and so on), not ours.
I wouldn’t put it quite that strictly myself, but I do lean the same way. We’re flooded with AI images these days that are simply fakes, and a lot of people can’t tell the difference. Good products deserve good photos – it also says something about how much a manufacturer cares.
I’m usually not a fan of EU (over)regulation, but this one I can actually get behind. That said, I hope @dan and the team find a way that also works for beginners and part-time web designers – marking a handful of images shouldn’t turn into a project of its own. Rules like this can be a pain in the a.., but I’d rather see it as a chance to make Elements a little better than as a threat.
Either way, good to see the topic getting some attention
I reckon the horse has long since bolted. I don’t think it is any longer possible to believe that any image you look at or any video that you watch is not AI generated or AI edited. Not sure a watermark will make any difference.
(Made with AI)