Explanations can take several forms: a breakdown of which requirements a candidate met, quotes from the CV or interview that support each rating, or a note on which factors weighed most in an overall score. A useful explanation is specific to the candidate and written so that a recruiter without technical training can follow it. A generic statement that a model 'found a strong match' does not count.
When a tool influences who gets an interview, recruiters need to be able to tell whether its reasoning is sound, and candidates may ask how an outcome was reached. Laws such as GDPR and the EU AI Act place weight on transparency and meaningful human review, which is difficult without clear reasons. Explainability also helps teams spot criteria that drift away from the actual job.