A job-related screening rubric you apply, not the AI
Resume screening is the second-most-common AI use in recruiting (44%) and also the most legally loaded. Letting a chatbot rank or reject candidates from their resumes is exactly what triggers bias-audit laws and disparate-impact liability. The safe, useful version flips it: have AI build a job-related scoring rubric from the role and give you a consistent way to record your own assessments — you do the judging.
You are helping a recruiter build a fair, job-related screening rubric for a role. Important boundary: you design the rubric; you do NOT evaluate, rank, score, or reject any actual candidate, and I will not paste resumes or candidate data into this chat. The role's real requirements and must-have skills: {{job_requirements}} How I currently decide who moves forward — my rough criteria: {{screening_criteria}} Produce: 1. A screening rubric that turns each must-have into an observable, job-related signal ("has shipped X" rather than "good communicator"), with a simple pass / maybe / no-signal scale and a weight showing which requirements matter most. 2. A short, consistent template I can use to write down why each candidate did or did not meet each criterion, so my own notes stay structured and comparable. 3. A "watch out" list: criteria in my rough notes that are risky proxies for protected characteristics (graduation year, employment gaps, "culture fit," school prestige) and a job-related signal to use instead. Rules: build the rubric only from the requirements I gave you — do not invent must-haves. Do not offer to screen resumes or rate candidates for me; the rubric is a tool I apply by hand. Keep it plain and auditable.
Fill in your details and the prompt updates live — then copy.
Screening rubric (weighted) - Owns paid-social budget (weight: high) — signal: has managed a stated monthly spend; pass / maybe / no-signal. - Hands-on Meta Ads Manager (high) — signal: names campaigns or optimizations they ran, not just "familiar with." - Can report on ROAS (medium) — signal: cites a metric they moved. Watch out (proxies to drop): - "No job-hopping" — can screen out caregivers and younger workers; instead look for evidence they shipped results in each role. - "Gets our culture" — vague and bias-prone; replace with a concrete work-style signal you can point to.
The full workflow
- Turn the job's real must-haves into the rubric — never paste a candidate's resume or PII to do it.
- Replace every proxy the "watch out" list flags with a job-related signal.
- Apply the rubric to real candidates yourself, writing notes in the template.
- Keep the completed rubrics — consistent, auditable notes are your defense if a decision is ever challenged.
Watch out for
If an AI tool scores, ranks, or filters candidates, it is likely an Automated Employment Decision Tool under NYC Local Law 144, which requires an independent annual bias audit, a public results summary, and 10 business days' notice to candidates. A consumer chatbot meets none of that — keep AI on rubric-building and let a human do the scoring.
Under Title VII you and the employer are liable for disparate impact even if a vendor's tool caused it. The EEOC removed its AI technical-assistance documents in January 2025, but the underlying law and state rules (NYC, Illinois, Colorado) still apply and are expanding.
Never paste resumes, names, or contact details into a consumer AI tool — that is candidate PII under GDPR/CCPA. Build the rubric from the job's requirements, not from real applicants.
Where this comes from
Every use case on this site is grounded in real reports from working recruiters — not invented by us.