Temperature · An LLM’s randomness setting
Temperature is a setting that controls how much randomness goes into an LLM’s answers. At 0, the model almost always picks the most likely token; the higher it goes, the more it picks less likely tokens, and the more answers vary from one run to the next.
Standalone definition · 47 words
How temperature changes an LLM’s answers
Temperature reshapes the probability distribution before the next token is sampled. Low values sharpen the top tokens; high values flatten the distribution. Metehan Yeşilyurt notes that 0 gives the most probable answers and that consumer assistants likely run around 0.7.
Why temperature matters for GEO tracking
With a non-zero temperature, the same prompt can mention different brands. One screenshot proves nothing: you need to repeat prompts and think in appearance frequency.
How to account for temperature in your tests
Run each prompt several times, record the appearance rate rather than a position, and also test at low temperature through the API to see the model’s “default” pick.
Frequently asked questions about LLM temperature
Can you set the temperature in ChatGPT?
Not in the app, but yes through the API for most models.
Does temperature 0 always give the same answer?
Almost, but not always exactly, due to server-side computation details.
Terms related to LLM temperature
Further reading on LLM temperature
- Best GEO tools 2026: top 10
- Alternatives to GetMint for AI tracking
- ← Back to the GEO glossary (150 terms)