Controls how the AI reads each application. Changes apply to future scans — existing applications keep the assessment they already have unless you re-run them individually.
How to judge, for every role. This goes into the model's instructions verbatim. Leave it empty to use the built-in default.
Two things aren't editable here and always apply: the AI ignores any instructions written inside an application, and it never comments on or scores a person's name, age, gender, nationality, ethnicity or photo.
The three buckets you sort by in the applications list.
Extra things you want weighed for a specific role that aren't in the public job description. Assessed exactly like the published criteria, and shown in the scan alongside them.