Refers to the reduction in variety, originality, perspectives, or approaches that can occur when artificial intelligence (AI) systems repeatedly generate, recommend, or influence content based on similar patterns, training data, prompts, or other AI-generated outputs. Research suggests that AI assistance may improve the quality or productivity of individual outputs across many users, making those outputs more alike; but the result can be a paradox: individual work may improve while collective creativity and diversity of thought may decline.

Diversity collapse may have many implications. As people increasingly rely on the same AI systems for writing, research, decision-making, problem solving, design, and idea generation, organizations and societies may risk producing more “standardized” ways of thinking and communicating. This scenario highlights the continuing importance of human judgment, lived experience, varied perspectives, human-created (original) information, and exposure to ideas outside AI-generated patterns.

“Model collapse” is primarily about deterioration in AI models, particularly when systems train recursively on synthetic data; “diversity collapse” is broader and can describe the narrowing of outputs and ideas even when the outputs themselves remain high quality.