Recruiting is a volume writing job dressed up as a people job. One requisition means a job post, three sourcing strings, a first-touch message, two follow-ups, an interview guide, a scorecard, and a dozen status emails. A chat assistant drafts all of it from one good intake call.
The bigger shift is on the candidate side. Applicants use the same tools to write resumes, answer screening questions, and rehearse interviews, so keyword matching means less than it did. Structured interviews, work samples, and reference checks matter more, and AI can help design those too.
The line you cannot cross: letting a model rank, reject, or score people without the audit, notice, and human review that New York City's Local Law 144 and Illinois' Human Rights Act amendments now require. Drafting is safe territory. Deciding is not.
Quick wins this week
- Turn a 15-minute intake call transcript into a job post, a must-have list, and five screening questions, then have the hiring manager mark up the must-haves.
- Ask for three Boolean strings for the same role at different strictness levels, test each in LinkedIn Recruiter or your ATS, and keep the one that returns a reviewable page.
- Rewrite your first-touch message so the opening line references something specific in the candidate's public profile, with a hard cap of 80 words.
- Convert your interview questions into a structured guide with a 1-to-4 rubric per question so every interviewer scores the same thing.
What AI can do for recruiters, task by task
Intake call to job post and scorecard
Record the intake call with consent, paste the transcript, and ask for a job post, a must-have versus nice-to-have list, and a scorecard with observable criteria. Check that every requirement is something you would actually screen on; models pad lists with credentials nobody mentioned.
Sourcing strings and talent mapping
Give the role, seniority, target companies, and location, and ask for Boolean strings with title synonyms and exclusions, plus adjacent companies to map. Test every string, and expect the company list to include firms that have shrunk or been acquired.
Personalized outreach at scale
Paste two of your best messages as voice samples, the candidate's public headline, and why this job fits. Ask for a message under 80 words with one specific hook. Read every fact before sending; the model will invent shared connections that never existed.
Structured interview guides and phone screens
Give the scorecard competencies and ask for behavioral questions, follow-up probes, and a 1-to-4 rubric with anchors. Then ask it to flag any question that touches protected characteristics or asks about disability before an offer. Managers adopt guides that fit on one page.
Debrief synthesis from interview notes
With candidate consent for recording, feed interviewer notes or transcripts and ask for evidence per competency, where interviewers disagree, and unanswered questions. Never ask for a hire or no-hire recommendation; the decision and its reasoning must be the panel's.
Pipeline updates and candidate communications
Paste aggregated stage counts (no names) and your notes, and ask for the weekly hiring-manager update: what moved, what is stuck, and what you need by Friday. Reuse the session for rejection emails that are kind and specific about the stage but never state a reason.
Prompts for recruiters
Replace the bracketed placeholders, paste into any chat assistant, and iterate on the result.
Intake call to job post and scorecard
You are a senior recruiter turning an intake call into hiring documents. From the transcript below, produce: (1) a job post under 400 words with no gendered or age-coded wording, (2) must-haves limited to what the manager said is required, with nice-to-haves separate, and (3) a scorecard of 4 to 6 competencies with observable criteria. Mark anything inferred rather than heard with [ASSUMED]. Context: [COMPANY, TEAM, AND WORK MODEL] Transcript: [PASTE THE INTAKE TRANSCRIPT]
Tip: Send the [ASSUMED] items back to the hiring manager as questions.
Boolean strings at three strictness levels
You are a sourcing specialist. Write three Boolean search strings for [JOB TITLE] at [SENIORITY] in [LOCATION]: one strict, one balanced, one broad. Include title synonyms, the key skills [LIST 3 TO 5 SKILLS], and exclusions for recruiters and job posts. Format them for [LINKEDIN RECRUITER OR YOUR ATS]. Then list ten companies likely to employ this profile, one line each, flagging any that may no longer exist under that name.
Tip: Run all three; keep the one that returns a page you can actually review.
First-touch message under 80 words
You are writing recruiting outreach in my voice. Two messages I have sent that got replies: [PASTE TWO SAMPLE MESSAGES]. Candidate's public headline: [PASTE HEADLINE]. Role and why it fits them: [ROLE AND FIT]. Write one message under 80 words: a specific first line about their work, one sentence on the role, and a low-pressure ask for a 15-minute call. Do not invent shared history, mutual contacts, or achievements.
Tip: Give real sample messages; without them the output sounds like every other template.
Structured interview guide with rubric
You are an industrial-organizational psychologist helping a recruiter build a structured interview. For the role [JOB TITLE] and these competencies [LIST COMPETENCIES], write two behavioral questions per competency, two follow-up probes each, and a 1-to-4 rubric with a one-sentence anchor for each score. Then flag any question touching protected characteristics, disability, family status, or salary history, and suggest a replacement. Keep the guide to two pages.
Tip: Salary-history questions are banned in many states and cities; keep that flag on.
Debrief synthesis by competency
You are assisting a hiring panel debrief. Using only the interviewer notes below, produce for each competency in [LIST COMPETENCIES]: the evidence cited, which interviewer cited it, and whether interviewers agree or conflict. Then list questions no interviewer asked that the scorecard requires. Do not recommend hire or no-hire, score the candidate, or speculate beyond the notes. Candidate: [CANDIDATE ID, NOT NAME] Notes: [PASTE INTERVIEWER NOTES]
Tip: Use a candidate ID rather than a name so the notes stay anonymized.
Want a prompt for something else? Use the Prompt Builder or browse hiring and management prompts.
Skills to build
Running a real intake call
Why: AI amplifies whatever you feed it, and a vague intake produces a polished post for the wrong job.
How: Use a fixed 12-question intake script covering must-haves, deal-breakers, comp range, and why the last person left. Record with consent and paste the transcript.
Structured interviewing and rubrics
Why: Structured interviews are the best defense against AI-coached candidates and bias, and they make debrief synthesis meaningful.
How: Build one guide with a rubric for each of your top three roles, and get hiring managers to score independently before the debrief.
Counterfactual bias checks
Why: A model can favor certain names, schools, or resume gaps in ways you will never see on a single output.
How: Run the same resume or message through twice with a swapped name, gender, or graduation year and compare. If anything changes, stop using the tool for that step.
Reading AEDT rules well enough to vet vendors
Why: The ATS ranking or video-interview tool you turn on may be an automated employment decision tool under NYC law, with the notice and audit duties on the employer.
How: Ask vendors for their latest bias-audit summary, what the tool automates, and how candidates are notified. File the answers with the requisition.
Verifying claims in an AI-assisted market
Why: AI-written resumes and coached answers make paper screening weak, and fraudulent remote candidates using altered video are a documented problem.
How: Add a short live work sample, ask for specifics on one project, verify identity for remote hires, and call references before the offer.
Tools worth knowing
ChatGPT
The general-purpose AI assistant most of your coworkers already use.
Claude
A careful writing and analysis assistant that shines on long documents.
Microsoft Copilot
AI inside Word, Excel, Outlook, and Teams, with your company's data protections.
Gemini
Google's assistant, strongest when your work already lives in Google Workspace.
Otter.ai
Meeting transcription and summaries that follow you into Zoom, Teams, and Meet.
Fireflies.ai
A meeting note-taker with deep integrations, search across calls, and an AI assistant.
Cautions for recruiters
Title VII, the ADA, and the ADEA cover every step of selection, and disparate impact from an AI ranking is the employer's liability even when a vendor built the model; the Mobley v. Workday litigation has tested vendor liability since 2023. Ask vendors for adverse-impact analyses, keep a recruiter in every screen-out decision, and check outcomes by protected group using the Uniform Guidelines' four-fifths rule as an early warning.
New York City requires an annual independent bias audit, a published summary, and notice to candidates at least ten business days before an automated employment decision tool is used on them. Illinois' AI Video Interview Act requires consent, explanation, and deletion on request for AI-analyzed video interviews, and its Human Rights Act amendments (effective January 1, 2026) require notice when AI informs hiring and prohibit using zip codes as a proxy for protected classes. Colorado's AI Act and the EU AI Act both treat hiring tools as high-risk; check Colorado's current effective date. Confirm with counsel before enabling scoring.
Resumes are personal data. California's CPRA covers applicants, GDPR covers EU candidates, and both carry retention and access duties. Do not paste full resumes into a consumer AI tool; use an approved enterprise tool, strip names, addresses, and graduation years first, and record interviews only with consent, which several states require from every party.
Models invent shared connections, misstate employers, and overclaim in debrief summaries, so read every fact before it reaches a candidate or hiring manager. On the other side, AI-written applications and deepfake video interviews are real; verify with live work samples, identity checks for remote hires, and references.
Your 30-day plan
- Week 1: Confirm your approved tool and what you may paste. Inventory every AI step in your funnel (ATS ranking, assessments, video interviews) and ask vendors about bias audits and candidate notices.
- Week 1: Build the intake-to-post prompt and run it on your next two requisitions; note what the hiring manager changed.
- Week 2: Rewrite your outreach with voice samples and compare reply rates against the old template for two weeks.
- Week 3: Build structured guides with rubrics for your top three roles, then pilot debrief synthesis, with consent, on one search.
- Week 4: Write down your personal rules (what you paste, what AI never decides), automate the weekly hiring-manager update, and tally hours saved.
Frequently asked questions
Will AI replace recruiters?
Can I use ChatGPT to screen resumes?
Do I have to tell candidates I'm using AI?
Is it okay to record interviews and have AI summarize them?
Terms used on this page
Related roles
- AI for HR ProfessionalsFirst drafts of policies, announcements, and survey analysis in minutes, with employee data and employment decisions kept where the law and your judgment say they belong.
- AI for Sales ProfessionalsSpend your hours in conversations instead of research tabs and CRM fields. Here is how sellers use AI for prep, follow-up, and proposals, and where the legal lines are on outreach and recording.
- AI for MarketersDraft campaigns faster, mine customer feedback for messaging, and turn reporting into a ten-minute job, without drifting off-brand or into a CAN-SPAM problem.
- AI for Small Business OwnersAI answers the one-star review calmly, drafts the month of posts, turns your voice memo into an SOP, and preps the questions for your CPA. You still make the calls, sign the checks, and own what goes out under your name.