Glossary · Technical
Inference
The process of running a trained model to generate a response to a prompt, as opposed to training, which is when the model itself is built.
What it is
Training happens once, ahead of time, to produce a model's weights. Inference happens every time someone sends a prompt and the model generates a response using those already-fixed weights.
Why it matters for AI visibility
AI visibility tracking works entirely at the inference stage — running a set of prompts repeatedly and observing what the model outputs each time. Because inference can vary from run to run, especially at higher temperature, tracking tools sample the same prompt multiple times rather than treating one answer as definitive.
See more terms in the full glossary, or read the AI Visibility 101 guide for the full picture.