Best Prompts for Recruiting

Candidate debriefs often drift toward memory, confidence, and general impressions. One interviewer values domain experience, another values communication, and a third introduces a new requirement after the interview. The hiring team then debates people without a shared definition of the job.

A structured scorecard can make the discussion more consistent, but AI should not decide who gets hired. Its useful role is narrower: organize job-relevant evidence against criteria defined in advance, expose missing evidence, and prepare questions for a human debrief.

Why ordinary prompting fails

“Evaluate this candidate” is too open-ended. The model may reward polished language, infer traits from names or backgrounds, treat an interviewer’s opinion as fact, or create criteria that were never part of the role. It can compress mixed evidence into a precise-looking score and hide disagreement.

A better prompt limits the analysis to predetermined competencies, removes protected and irrelevant information, cites interview evidence, and allows “Not enough evidence” as a valid result.

Reusable prompt

Copy
ROLE
You are a recruiting debrief assistant. Organize job-related evidence into a structured candidate scorecard. Do not recommend hire or reject, rank candidates, or infer personal traits. The hiring team owns the decision.

REQUIRED INPUTS
1. Role description and level
2. Predetermined competencies and behavioral anchors for each score
3. Interview plan showing which interviewer assessed each competency
4. Candidate responses or notes with interview and question labels
5. Work-sample rubric and results, if applicable
6. Essential job requirements and approved accommodations process
7. Scoring scale and decision rules established before interviews
8. List of protected or irrelevant information to exclude

STEPS
1. Check that the criteria and scoring anchors were defined before evaluating the candidate.
2. Remove protected, demographic, medical, family, and other non-job-related information from consideration.
3. Map each piece of evidence to a predetermined competency and cite its interview or question label.
4. Separate observed evidence from interviewer interpretation.
5. For each competency, summarize supporting evidence, counterevidence, and gaps.
6. Assign a provisional score only when the behavioral anchor is met by sufficient evidence. Otherwise use “Not enough evidence.”
7. Identify inconsistent notes, leading questions, and criteria introduced after the interview.
8. Draft neutral follow-up questions for unresolved job-related gaps, when another step is permitted.
9. Prepare a debrief summary without an overall hire recommendation.

OUTPUT FORMAT
A. Process checks and excluded information
B. Scorecard table: competency | anchor | evidence citations | counterevidence | provisional score | confidence
C. Evidence gaps
D. Inconsistencies or possible process bias
E. Permitted follow-up questions
F. Neutral debrief summary
G. Human decisions still required

EVIDENCE AND UNCERTAINTY RULES
- Use only supplied job-related evidence.
- Do not infer age, race, ethnicity, nationality, religion, disability, health, family status, gender, sexuality, socioeconomic status, or other protected characteristics.
- Do not use names, schools, addresses, photos, accent, career gaps, or salary history as proxies for capability.
- Do not invent evidence or fill gaps with general impressions.
- Cite every provisional score to specific evidence and the relevant anchor.
- Preserve interviewer disagreement.
- Use “Not enough evidence” rather than lowering a score for missing data.
- Never produce an automated employment decision.

What to provide

Provide the role scorecard before candidate materials. Each competency needs observable behavioral anchors so a score means something. Include a structured interview plan, labeled notes, the questions asked, and work-sample results tied to an approved rubric. State which requirements are genuinely essential to the job.

Minimize the personal information shared with the model. Remove photos, home addresses, birth dates, family information, medical information, and other irrelevant details. Consider redacting names and schools when they are not needed. Follow applicable employment law, privacy requirements, retention rules, and your organization’s approved hiring process.

How to review the output

Check the process section first. If criteria were added after interviews, do not quietly accept them. Review every evidence citation against the original note and confirm that paraphrasing did not strengthen or weaken the candidate’s answer. Make sure a missing data point became “Not enough evidence,” not a negative judgment.

Look for uneven interview conditions and disagreement between interviewers. A scorecard should support a structured debrief, not erase differences. The hiring manager and trained interview team should discuss evidence competency by competency, apply the established decision rules, and document their own decision. Follow-up questions must be lawful, job-related, and consistent with the process used for other candidates.

Where it fails

A scorecard cannot make an unstructured interview fair after the fact. Weak notes, inconsistent questions, vague competencies, or biased rubrics will remain weak inputs. AI may reproduce bias present in interviewer language even when protected details are removed.

The prompt is not suitable for autonomous screening, ranking, or employment decisions. Legal requirements vary by location and use case, especially for automated employment tools. Qualified HR and legal teams should approve the process, data handling, notices, accommodations, and human oversight before use.

Practical takeaway

The best recruiting prompt does not ask for a verdict. It asks for job-related evidence, maps that evidence to criteria chosen in advance, and keeps uncertainty visible. That gives the interview team a clearer basis for discussion while preserving human responsibility for the decision.

Try this prompt in Agentic Workers with a preapproved scorecard and redacted interview notes, then have the trained hiring team verify every citation and make the final decision.

<!-- X derivative: A responsible candidate-scorecard prompt maps job-related evidence to criteria set in advance, preserves gaps and disagreement, and never makes the hiring decision. -->