Primary Bias
An AI model's ungrounded confidence in an entity, formed during training and present before any retrieval; in AI search it is the largest single factor in whether a source is selected.
Primary bias is an AI model's ungrounded confidence in an entity, formed during training and present before any search or retrieval happens. It is the model's pre-judgement, baked into its weights, and in AI search it is the largest single factor in whether a source is selected.
It is slow to change because it lives in the training data, and it is measured through its effect on the Selection Rate. Full article: Primary Bias. Its post-retrieval counterpart is secondary bias, which includes grounding bias. Related: AI Visibility.