Image Rights: Stock, Team Photos and AI Pictures

Putting someone else's photo on your website without permission touches two separate rights at once. First, the right to make a work available so that the public can access it from places and at times of their choosing (German Copyright Act, Section 19a). Second, the author's right to recognition of their authorship and to decide whether and how the work carries an author's designation (German Copyright Act, Section 13). Those are two independent claims. That is why a demand can still arrive long after the image has been deleted: injunctive relief, damages and the costs of a warning letter do not depend on whether anyone acted deliberately.

In practice, a small number of highly recurrent situations lead to a claim. Once you know them, you tend to recognise your own image library within minutes.

One widespread misconception concerns cost. The German Copyright Act caps the value in dispute for the legal fees of a first justified warning letter at 1,000 euros (German Copyright Act, Section 97a) — but expressly only towards natural persons who do not use the work for their commercial or self-employed professional activity. A company website does not fall under that. For businesses the cap therefore does not apply, and the value in dispute follows the economic interest in the individual case.

Buying an image is not a purchase. You acquire a right of use within a defined scope, and that scope sits in the licence text, not on the invoice. Anyone who invests five minutes and saves the licence text alongside the download has settled the proof for the entire time the image is in service. Seven questions decide almost every later dispute.

Of all these points, the duty to credit is the one most often underestimated. It follows not only from the licence but from the author's personality right in the work (German Copyright Act, Section 13). Even under a fully paid licence the agreed designation remains owed, and its absence is regularly asserted in practice as an independent infringement — often with a surcharge on the customary fee. The effort needed to avoid this amounts to one line of text per image.

The second typical breach happens on re-use. Experience shows that a motif licensed for the website migrates within a few months into a social profile, into a quotation PDF and onto a stand. Each of those uses is a separate matter. The least friction comes from a plain internal rule: before an image enters a new channel, the licence text is read once more. If it does not cover the channel, the licence is extended or a different motif is chosen.

As soon as people are identifiable in a photo, a second body of rules joins copyright. The German Act on the Protection of Copyright in Works of Art and Photography sets out the principle that portraits may only be disseminated or publicly displayed with the consent of the person portrayed (Act on Copyright in Works of Art, Section 22). The same provision contains two subtleties that are rarely known: if the person received remuneration for being portrayed, consent is deemed granted in case of doubt. And after their death, dissemination requires the consent of their relatives for a period of ten years (Act on Copyright in Works of Art, Section 22).

The statute knows exceptions, for instance for people appearing incidentally beside a landscape or for participants in assemblies and processions. In advertising these exceptions rarely carry. A photo on a company website serves the outward presentation of a business, and courts read such exceptions narrowly in that setting. For everyday practice the rule is simple: whoever is identifiable gets asked — and the answer gets recorded.

The second group of people is frequently overlooked: customers. Site photos show clients, dining rooms show guests, practice and office waiting areas show people. Here too consent is needed where the person is identifiable. The workable answer is to plan the shoot so the question does not arise: appointments outside opening hours, deliberately chosen framing, perspectives from behind. What additionally needs clarifying for reference projects is covered in the article on client references and logos.

With image generators the legal position becomes surprisingly simple in one respect and more complicated in another. It is simple regarding the output: German copyright protects personal intellectual creations. An image produced solely by entering a text prompt regularly does not qualify, because the concrete design comes from the machine. Nor does the neighbouring right in photographs help, since it presupposes a human photographer (German Copyright Act, Section 72). The practical consequence cuts both ways: you pay no licence fee — but you can barely stop others from using the same motif.

It gets more complicated with everything that goes into the image and everything that remains recognisable within it. The provider's terms of use govern what you may do with the output, and they frequently distinguish private from commercial use. Source images you upload bring their own rights along: someone else's photo used as a reference remains someone else's photo. And a model can produce elements that touch third-party rights without you intending it.

A short log per motif answers most of these questions in advance. Record the provider, the date, the full prompt, any source images and the terms of use applying at that moment. It takes under a minute and later answers three questions at once: where did this image come from, was commercial use permitted, and does it need a disclosure notice? How generative tools fit sensibly into everyday website work is covered in the article on what is possible when building a website with AI.

Image files carry fields in which origin and conditions travel along. This is the only proof still present when the file gets separated from its context — during a move to a new system, when sending it to a print shop, when passing it on to a colleague. The fields are quick to fill and reachable in every common image application. The same standard applies to the description as to the rest of the page content: one clear sentence without marketing language, of the kind set out in the article on writing website copy customers actually read.

Two caveats matter. First, metadata does not necessarily survive image optimisation: tools that shrink files and convert them into modern formats often strip the fields because they cost bytes. Relying on them alone means losing the proof at the very moment the file goes live. Which formats and sizes make sense in the first place is covered in the article on image formats and alt text. Second, camera location data rarely belongs on the web — for shots taken in a home office it holds a private address.

The cause of most image rights problems is not legal but organisational: the image, the proof and the point of use sit in three different places. XICflow starts exactly there. Image motifs are created per website inside the project itself — matched to the sector, the colour world and the copy on the page — and stored in the media library. Every object carries an alt text, a title and an origin note recording the source, the condition and any AI disclosure.

For images you bring along — your own shots, licensed motifs, photos from a photographer — nothing changes about the legal framework. What changes is the location: the note on licence or consent sits at the image and can be searched, instead of living in a folder on a computer that nobody opens once a colleague has left. What this looks like in a finished website is shown by the example sites in the demos; the scope of the media management is described in the overview of features.