From Slop to Slop: On the Eerie Afterlives of (Generative) Images

Generative AI models are trained on billions of images systematically scraped from the Internet, including personal photos and other traces of everyday life. Many of these images were never meant to last, or at least no one was ever asked if they wanted them to be immortal. Yet they persist, absorbed into ever-growing training data sets. GenAI recycles these images into new outputs, giving them a largely involuntary afterlife. Drawing on Mark Fisher's concept of the "eerie" and the field of critical data (set) studies, this article traces the flow of images through GenAI models. By analysing this process and its sloppy output, I argue that generative AI carries the material traces of people's digital lives, repurposed without consent and given an infrastructural, messy immortality. The eeriness of slop makes visible a form of afterlife that is neither intentional nor preservative, but infrastructural: images transformed beyond recognition, yet never fully gone.