Image integrity in the age of generative AI

Tuesday, 16 September 2025, at 2pm UK time

Establishing the authenticity of images is not a new issue. In 1840, Hippolyte Bayard staged a photograph called ‘Self-Portrait as a Drowned Man’, in protest to the lack of recognition he received as one of the principal inventors of photography. Bayard’s photograph may have been part conceptual art and part publicity stunt, with no intention to permanently deceive, but since then, manipulation and faking of images has been misused many times in efforts to substantiate false claims, commit fraud, and manipulate public perception.

Inappropriate use of images used in scientific articles is now more visible thanks to the works of so-called ‘science sleuths’ like Elisabeth Bik, Kevin Patrick, David Bimler and Jana Christopher. These sleuths often share their findings on platforms like PubPeer, making the platform a useful tool for scientists and publishers alike. More and more studies are being published on the topic, with findings ranging from 4% up to 26%, depending on the body of literature that has been investigated. According to the US Office of Research Integrity, nearly 70% of all fraud cases they investigate are about images, and that’s just the ones that get noticed.

Times are changing, though. It’s getting increasingly more difficult to determine whether an image is authentic or not. And while we’ve all seen examples of ‘AI slop’ on the internet, where it’s still fairly easy to spot fake photographs from telltale signs like unusual or inconsistent details, it’s much harder to detect whether a microscope image or western blot has been altered or entirely faked. Even without the benefit of AI tools, technically savvy fraudsters are able to use software to back up false claims and findings by moving lines in western blots, copying and distorting parts of microscopy images, or even reusing images published elsewhere and passing them off as new.

In this webinar, attendees learnt about the importance of image integrity to safeguarding the scholarly record. The speakers provided an overview of some of the many approaches and techniques that currently exist to push back against fake research in general and manipulated and falsified images specifically. Finally, they looked to the future to explore some of the technologies that are being developed -or already exist in other domains- that may well become future image integrity requirements for publishers and editors.