The Generative Bubble

The generative bubble is the academic name given by researchers at the London School of Economics in 2025 to the cognitive-filter phenomenon in AI production environments. The researchers' precise formulation — "in the generative bubble, users are filtered, limited, or restricted by themselves alone" — captures what distinguishes this bubble from Pariser's original. The content filter bubble was imposed by external curation. The generative bubble is co-created: the user's prompts carry the signature of her cognitive architecture, and the model's statistical tendencies respond to that signature by generating outputs aligned with it. The alignment feels like understanding; its structural reality is confinement. The bubble is constituted not by a third party choosing what to show but by the feedback loop between the user's existing patterns and the model's probabilistic generation.