Model Drift
Definition
Model drift is change in an AI system’s outputs over time for the same prompts, caused by updates, tooling, or data shifts.
By Vinespire Editorial Team, Editorial ·
This term is part of the full AI search glossary.
Full definition
Generative products ship new model versions, system prompts, browsing tools, and safety layers. A prompt battery that ranked your brand last month may omit it after a silent update—or suddenly cite you more often—without any change on your site.
Drift differs from content decay on your properties: the external system moved. Detect it by versioning runs, noting product release notes when available, and separating site deploys from model change windows in analysis.
GEO programs need continuous monitoring, not one-off audits, because drift is normal. See Visibility Probe, Prompt Battery, and Fact Consistency.