Glossary · Technical
Temperature
A setting that controls how random or predictable a model's output is — low values produce focused, repeatable answers; high values produce more varied ones.
How it works
Temperature scales the probability distribution the model samples from when choosing its next word. At low temperature it almost always picks the most likely next word, producing consistent output; at higher temperature it more readily picks less likely words, producing more varied phrasing and, sometimes, different conclusions.
Why it matters for AI visibility
At low temperature, the same prompt tends to surface the same top brands run after run. At higher temperature, the exact same question can yield a different set of mentioned brands on different runs, which is one reason visibility tracking tools sample a prompt many times rather than trusting a single answer.
See more terms in the full glossary, or read the AI Visibility 101 guide for the full picture.