From 2 August 2026, AI-staged property photos in the EU must be labelled as artificial

The label turns on one judgement. According to a dev.to writeup on the rules, an estate agency has to ask whether AI generated the image or substantially altered it. Whether the image was edited at all is not the test ^[4].

"Substantially" is where the build gets awkward. Its examples are virtual furniture, digitally created landscaping and major AI changes to a room's appearance ^[5]. Beyond those cases, the writeup says the boundary between routine editing and substantial AI alteration will need careful assessment as implementation practices mature ^[6]. Put plainly, the line has not been drawn yet. A pipeline therefore stores a judgement that may later be revised. I would record each one with the name of the person who made it.

The writeup's process is declaration at intake, with disclosure built into content production ^[18]. When a photographer, marketer or external agency submits an image, the team works through four documented questions. Was generative AI used? What did it change? Is the change substantial? Where will the image appear ^[12]? The team should also record the AI tool and the nature of the change whenever staff, photographers, agencies or contractors supply media ^[11]. In a listings system those answers are required fields on the upload form. An asset with any of them blank should not reach the publish queue. The writeup does not describe automated detection or name a machine-readable format.

Two labels come out of that record. The visible one has to be clear enough for a prospective buyer or tenant to notice and understand ^[7]. The writeup argues that a plain caption such as "AI-staged image" is easier to apply consistently than a disclaimer buried in listing copy ^[13]. I agree. A caption is a template element with a fixed position. A sentence inside listing copy gets rewritten every time someone edits the description. The machine-readable label is harder to guarantee. Businesses are asked to preserve it only where their tools and the publishing channel support it ^[9].

Syndication adds a second problem. One property can appear on the agency's own website, on third-party portals and in paid social campaigns, and each has different publishing controls ^[16]. The writeup recommends one baseline process for labels and records across EU markets, so the same listing does not go out with different levels of transparency ^[15]. Its checklist extends the template review to email campaigns and paid advertising ^[17]. I think each channel's template should render the caption from the asset record. Then the decision is made once at intake and every channel reads it from the same field.