Each generation smooths over rare patterns and outliers in favor of the statistical average, so diversity and accuracy erode over successive training rounds. As AI-generated content spreads across the web, it becomes harder for future models to train on data that wasn't itself touched by a model, which is the practical worry behind the term.