Temperature is a setting that controls how predictable an AI model's output is. Low temperature gives steadier, more repeatable answers; high temperature gives more varied and creative writing. Most chat apps set it for you, while APIs and some builder tools let you change it.
AItopic
2 minread
1related terms
In one line
A setting that controls how predictable an AI's output is. Low temperature gives steadier, repeatable answers; high temperature gives more varied writing. Most chat apps set it for you.
When a model writes, it picks each next token from a list of likely options. Temperature changes how it chooses. At a low setting it almost always takes the most likely option, so the same prompt gives similar answers each time. At a higher setting it more often picks less likely options, which gives fresher wording but also more drift, odd phrasing and a higher chance of mistakes.
In ChatGPT, Claude or Grok chat apps you usually cannot see the number. You meet it when you build with an API, use an automation platform or configure an AI feature inside recruiting software. If you cannot set it, you can still steer the result: ask for a fixed format and exact rules when you want consistency, and ask for several different options when you want variety.
Why it matters to recruiters
Recruiting tasks need different levels of predictability. Scoring resumes against a rubric, extracting skills or summarizing interview notes should give the same answer every time, so low temperature fits. Brainstorming subject lines or job ad hooks benefits from variety. If you are setting up an AI workflow in your ATS or automation tool, choosing the right setting cuts rework and odd outliers.
An example from a recruiting desk
One TA team builds an automation that reads each new application and tags must have skills. At a high temperature the same resume gets slightly different tags on two runs, which confuses hiring managers. They drop the temperature to near zero, add a fixed list of allowed tags, and the output becomes consistent enough to filter on.
Tips that help
Use low temperature for scoring, extraction, summaries and anything you filter or report on.
Use a higher setting only for brainstorming copy you will edit anyway.
Pair low temperature with a fixed output format for the steadiest results.
Test any change on the same ten inputs before rolling it out.
Mistakes to avoid
Raising temperature to fix weak output when the real problem is a vague prompt.
Expecting zero temperature to make a model factually correct; it only makes it more consistent.
Frequently asked questions
What temperature should I use for resume screening?
Use a low setting, close to zero where your tool allows it. Screening should be consistent and explainable, so you want the same resume to get the same result each time. Combine it with a clear rubric and a human review of decisions.
Can I change temperature in ChatGPT or Claude chat?
Normally not in the standard chat apps. The setting is exposed in developer APIs, playgrounds and some automation tools. In chat you get a similar effect by asking for strict formats when you want consistency, or several options when you want variety.
Does low temperature stop AI hallucinations?
No. Low temperature makes answers more repeatable, but a model can repeat the same wrong fact with confidence. Reduce errors by giving it the source material, asking it to quote from that material and checking claims before anything goes to a candidate or client.
