Temperature
Temperature is a setting that controls how random or deterministic an AI model's output is, with higher values producing more varied responses and lower values more predictable ones.
Temperature dials how much an AI model improvises. At low temperature the model picks the most likely next token almost every time, giving consistent, conservative answers; at high temperature it samples more freely, producing varied and creative — but less predictable — output.
It explains a phenomenon every AEO practitioner hits: ask an engine the same question twice and you can get different answers, including different cited sources. That run-to-run variation is partly temperature (plus other sampling and retrieval randomness), and it's why measurement has to be the adaptability pillar — sample a fixed prompt set repeatedly and watch the trend rather than trusting a single reading. Higher temperature can also raise the odds of a hallucination.
Example. Run "best CRM for small teams" through an engine five times and you may see your brand cited in three of them. That inconsistency is expected; tracking the rate across many runs, not one, is how you read your real visibility.
Relevant pillar
Related terms
- Large Language Model (LLM)A large language model is an AI system trained on vast amounts of text to predict and generate language, and is the engine that writes the answers in AI search.
- HallucinationA hallucination is when an AI model states something false or fabricated as if it were fact, usually because it generated from memory instead of grounding its answer in real sources.
- System PromptA system prompt is the hidden instruction that sets an AI assistant's behavior and rules before it sees the user's question, shaping how it answers and what it's allowed to do.