The Pope in a white designer down jacket. Donald Trump during an alleged arrest. Both images went viral, and neither was real. So-called deepfakes, created using artificial intelligence, are no longer a novelty. A study by the University of Leipzig now shows that as deepfakes become more widespread, so does mistrust of real images and videos.
Researchers at the University of Leipzig surveyed approximately 1,300 internet users in Germany in the summer of 2025. In 2022, a good three-quarters of respondents said they had never consciously seen a deepfake. In last year’s survey, that figure had dropped to only about 27 percent. Users encounter fake content primarily on Instagram, Facebook, YouTube, and TikTok.
“What was long considered a curiosity is now part of normal media consumption,” says Prof. Dr. Alexander Godulla, one of the two project leaders of the “Deepfake Project” at the Institute for Communication and Media Studies at the University of Leipzig.
What helps now – and what doesn’t
According to the researchers, anyone who wants to check whether an image or video is fake should pay attention to three things. First, question your gut feeling: Content that strongly confirms your own worldview deserves a second look. Second, look for technical details, such as transitions between the face and body, hairlines, or missing light reflections in the eyes. Third, check the source: Can the image or video be traced back to an original text?
“The most reliable way to verify something isn’t just looking at it, but doing further research,” says Hoffmann. Societal resilience against manipulation does not arise from blind mistrust. It requires “intellectual humility – the admission that I, too, can be mistaken.” And it requires strong institutions: credible journalism and media literacy education that begins early, especially for children and adolescents.
The challenge in the future will not only be to recognize fakes. It will be just as important to maintain trust in reliable information.
Original publication:
Bendahan Bitton, D., Godulla, A., & Hoffmann, C. P. (2026). Deepfakes and Synthetic Media in the (Digital) Public Sphere: Between Knowledge Gaps, Risk Perception, and Acceptance. University of Leipzig.
