PromptScoutWhat is Model Drift?
Model Drift is a change in a model’s behavior or output characteristics over time due to updates, data changes, or shifts in underlying systems. Model Drift can affect answer consistency, citation patterns, and reliability.
Quick definition
Model Drift means an AI model’s behavior changes over time.
How Model Drift works
- Model Drift can occur after model updates, fine-tuning, or changes in retrieval systems.
- Model Drift can change how a model answers the same prompt.
- Model Drift can change how sources are selected and ranked when retrieval is involved.
- Model Drift can be detected with prompt monitoring and prompt tracking.
Why Model Drift matters
Model Drift matters because monitoring baselines can become invalid when behavior changes.
Model Drift also affects AI visibility metrics because citation presence can shift without content changes.
Example use cases
- Detecting that a stable prompt now produces different definitions.
- Observing that citation in AI answers disappears after a system update.
- Comparing outputs month over month for the same prompt index.