For the cover image of my song about Swiss General John Sutter—who founded “New Helvetia” in the U.S. and was caught off guard by the California Gold Rush—I asked the AI to provide a public-domain historical image. The cover image I’ve now chosen for the song seems to do just that: it’s a view from a historical archive of the general on horseback in front of Fort Sacramento, which he founded. I actually had no doubts about the origin of the old engraving, but I wanted to be sure nonetheless.

But then the AI bot gave me a rather surprising answer: “This image was generated entirely from scratch by AI and does not come from a real historical source.” The image file contains secure data records indicating that the media was created and edited by Google LLC using AI tools. This type of data record is called C2PA Content Credentials. They securely embed the origin and editing history of a piece of content directly into the file to ensure transparency regarding the creation of digital media.”

But that also means it’s no longer as easy to recognize the fake nature of images just by looking at them from the outside—for example, when, during an art class at school, you used to notice that a person had been removed from a picture. And the tech companies are covering their bases because they know that copyright law could be one of their stumbling blocks. These companies—in this case, Google LLC—embed the image’s origin and editing history within the image itself, though of course this isn’t visible from the outside. This is of little help to us laypeople when we’re raising questions about an image in the context of media education.

The good news: As long as the content credentials weren't removed when the file was shared, they can be viewed using Adobe's official "Content Credentials Viewer": You can upload images there and view the stored information.

That may be reassuring for experts who want to research the provenance of such images. But in everyday life, it’s becoming increasingly difficult to distinguish between real depictions of reality and artificially created reality. After all, you can’t run every image through such a reader just to be sure. The problem is also so serious because it is no longer the work of a single forger. Rather, the information available to the AI is so vast that, naturally, many traces of real images and information flow into the newly constructed image of General Sutter. The lines between fake and real blur to such an extent that it is no longer possible to clearly distinguish between “false” and “true.” The creator of such an image ultimately generates a new form of reality that draws on facets of reality and their own imagination.

For us, this means that simply critiquing an image on a visual level is hardly effective anymore. Instead, source criticism—examining context, authorship, and intent—is becoming increasingly important and taking center stage. We can no longer trust the image itself; instead, we must examine the path it has taken to reach us. But even these paths are often so convoluted that we are still far from solving all the mysteries.