ChatGPT now has 900M+ weekly active users and 50M+ consumer paid subscribers as of February 2026 (TechCrunch). Inside recruiting, 27% of organizations use AI in their recruiting workflow per SHRM's State of AI in HR 2026 report, and recruiters are the most frequent AI users at 46% per iCIMS and Aptitude Research (April 2026 study, n=400+ US TA leaders). LinkedIn's Future of Recruiting 2025 report puts the productivity number more concretely: recruiters using GenAI save roughly 20% of their work week, equivalent to one full workday per week.
The catch: a bad prompt produces bad output, and most listicles on this topic are giving you prompts written by people who never tested them at scale. This piece is the opposite. The 21 prompts below are organized by recruiter task, written using a consistent prompt formula, and each one is paired with the specific recruiting outcome it is engineered to drive.
We also tell you where ChatGPT breaks (it does, often) and what to use instead. At Yander, we built the AI sourcing and outreach platform that does the things ChatGPT cannot, starting at $89 per month from yander.ai/pricing with no placement fees.
The 5-part prompt formula that gets ChatGPT to do recruiter-grade work
Most recruiter prompts fail because they skip structure. ChatGPT is a probability machine that picks the next likely token given everything it has seen. Vague input gets you vague output. Specific structured input gets you usable output.
Every prompt below uses the same five-part formula:
- Role: who ChatGPT is acting as ("You are a technical recruiter at a Series B SaaS company hiring backend engineers.")
- Context: the specific situation ("We use Python, FastAPI, PostgreSQL, and deploy on GCP. We have just lost our second senior engineer to a competitor.")
- Constraint: what to do and not do ("Generate 5 boolean search strings. Do not use the keyword 'rockstar.' Each string must include at least one current-employer filter.")
- Format: how to structure the output ("Return as a numbered list. After each string, add a one-sentence explanation of who it targets.")
- Tone: voice register ("Keep the explanation operator-direct; avoid salesy language.")
You can write a great prompt without all five, but missing one consistently weakens output. Skipping Role gets you generic answers. Skipping Format gets you walls of text. Skipping Constraint gets you ChatGPT inventing things you did not ask for.
The prompts below all follow this structure. Adapt the bracketed sections to your own role, stack, and stage.
Sourcing prompts
1. Boolean search strings for a specific stack
You are a senior technical recruiter sourcing on LinkedIn. I am hiring
a Senior Backend Engineer at a Series B SaaS company in San Francisco.
The stack is Python, FastAPI, PostgreSQL, Redis, deployed on GCP. We
prefer candidates who have worked at Series A-C startups (not FAANG).
Generate 5 boolean search strings I can paste into LinkedIn Recruiter
or a Google X-ray search. Each string must:
- Include "Senior" or "Staff" in title filter
- Include at least 3 of our stack keywords
- Exclude common false-positive titles ("Sales Engineer", "Solutions
Engineer", "Product Engineer")
- Include a current-employer filter favoring Series A-C SaaS
Return as a numbered list. After each string, add one sentence on
exactly who it targets and one sentence on the likely false-positive
risk. Operator-direct tone, no fluff.
2. ICP mapping for a niche role
You are a research analyst helping a recruiter define an ideal-candidate
profile. We are hiring a Staff ML Engineer for a 50-person AI-native
SaaS company building agentic workflow tooling. The role is hands-on
with PyTorch and JAX, with system-design ownership for our inference
infrastructure.
Generate a structured ICP brief with:
- 5 target companies where this person likely works today
- 3 target titles they currently hold
- 5 specific GitHub projects or conferences they likely contribute to
- 3 reasons they would consider moving in 2026
For each named company, conference, or project, only include it if
you can name a specific URL, repo, or event page (e.g.,
github.com/jax-ml/jax). If you cannot name a specific verifiable
source for an entry, replace it with the placeholder
"[research needed]". Do not invent.
Format as a markdown table. Avoid generic answers.
3. Google X-ray search for GitHub-active engineers
You are a sourcing specialist using Google site-restricted search to
find software engineers active on GitHub. I want to find engineers who
have shipped Rust + WebAssembly side projects in the last 12 months
and currently work at a US SaaS company.
Generate 3 Google X-ray search queries using site:github.com that:
- Filter for users with "github.com/[username]" profile URLs
- Include Rust and WebAssembly keywords
- Include "Senior" or "Staff" or "Principal" in their bio or profile
- Exclude common false positives (forks, tutorial repos, language docs)
Return as a numbered list with the exact search string and one sentence
on what each query is engineered to surface.
4. Title-permutation generator for hard-to-search roles
You are a senior recruiter. I am sourcing for a role where the title
varies wildly across companies. The role is essentially "engineering
leader of the platform team at a 100-300 person SaaS company." That
title appears as: Engineering Manager, Director of Platform, Head of
Infrastructure, Staff Engineer (platform), Principal Engineer
(platform), VP of Engineering at smaller companies.
Generate 12 title-permutation strings I can OR-combine into a single
LinkedIn search. Each entry should be the literal title text in quotes.
Format as a single OR-joined string ready to paste into LinkedIn
Recruiter title field.
5. ICP "no-go" list
You are a senior recruiter helping me tighten my sourcing pool by
filtering OUT poor-fit profiles. I am hiring for a senior product
engineer role at an early-stage YC startup.
Generate a checklist of 10 disqualifying signals that show in a
LinkedIn profile, with one sentence each on why this signal predicts
poor fit for an early-stage startup specifically.
Format as a numbered list. Be specific and unsentimental. If a signal
is contextual, say so.
For the deeper guide on running this sourcing layer at scale, our AI candidate sourcing tools comparison for 2026 maps the platforms that aggregate live candidate data across LinkedIn, GitHub, and other public signals.
Outreach prompts
ChatGPT can help you draft outreach faster, but quality at scale comes from automation and personalization data that ChatGPT alone cannot provide. LinkedIn's own published data: AI-Assisted Messages see a 44% higher acceptance rate and are accepted over 11% faster than non-AI messages (LinkedIn 2025 Hiring Release). The Hiring Assistant product reports 66% higher InMail acceptance than traditional sourcing (LinkedIn product page).
6. First-touch outreach (engineering hire)
You are an experienced engineering recruiter writing a personalized
first-touch outreach message to a passive senior engineer on LinkedIn.
About the candidate (paste in their profile bullets):
- [Current role + company + tenure]
- [One specific project or repo they have shipped]
- [One signal about what they care about, talk, blog post, etc.]
About my opening:
- Role: Senior Backend Engineer at [Company], Series B SaaS
- Stack: [3-5 specific items]
- Comp band: $[X] to $[Y] total
- Work arrangement: Remote / Hybrid in [city]
- The 90-day deliverable: [one specific problem they would own]
Write a 100-130 word message. First sentence references the specific
project I named. Avoid generic openers like "I came across your profile." The ask
is a 15-min call. Tone: respectful, direct, no salesy language. End
with my first name only.
7. Follow-up sequence for no-response
You are a recruiter writing a 3-message follow-up sequence for a
passive candidate who did not reply to my first outreach 7 days ago.
Original message context: [paste your first message]
Constraints for each follow-up:
- Follow-up 1 (day 4): 40-60 words, references a different specific
thing about the candidate, no "circling back" or "checking in"
phrasing
- Follow-up 2 (day 11): 30-40 words, single direct ask
- Follow-up 3 (day 21): 25-30 words, soft close offering to re-engage
in 6 months
Format as 3 numbered messages with the day label as the heading.
8. Personalization layer for an outreach template
You are a recruiter personalizing a templated outreach message at
scale. I am about to send the message below to 80 candidates. For each
candidate, I have one specific signal from their LinkedIn profile.
Template:
[paste your generic template]
Candidate signal:
[paste one specific signal, recent talk, recent role change, recent
repo activity, etc.]
Rewrite the FIRST SENTENCE of the template to specifically reference
the candidate signal. Keep the rest of the template unchanged.
Generate one rewrite per candidate signal.
Tone: respectful and specific. Avoid flattering phrases like "I love your work."
9. Hiring-manager-voice outreach (SOBO)
You are a hiring manager (not a recruiter) writing a personal outreach
to a passive candidate. The recruiter has identified the candidate as
a strong fit. I am the hiring manager, Director of Engineering at
[Company].
Candidate context: [paste candidate profile highlights]
Write a 100-word message from my perspective as the engineering
leader. Mention one specific technical challenge the candidate would
own. Be direct about why I personally think they are a fit (not
generic "we are impressed"). Ask for a 20-min call with me directly,
not a screen with the recruiter.
Tone: peer-to-peer technical. Avoid corporate recruiting voice.
10. Re-engagement message for a previously-silent candidate
You are a recruiter re-engaging a senior candidate I reached out to
6 months ago. They did not respond. Now we have a new role open and
an updated comp band.
Old role context: [brief]
New role context: [brief, what changed]
What I want from this message: a fresh opening with no reference to
my old unanswered outreach. The candidate should not feel "chased."
Write a 90-110 word message that opens with the new context, makes the
new role specific, and asks for a single 20-minute conversation.
For the full multi-channel outreach automation layer that actually executes these messages at volume, our recruitment automation software guide for 2026 compares the tools that sequence and track outbound at scale.
Job description prompts
JD writing is the single biggest AI use case in recruiting. Among organizations using AI to support recruiting efforts, 66% use it for writing job descriptions (SHRM 2025 Talent Trends).
11. Write a JD from a bullet-point intake
You are a senior recruiter writing a job description from a hiring
manager intake.
The intake notes:
- Role: [Title]
- Team: [Team and reporting structure]
- 90-day deliverables: [3 bullets]
- Required experience: [3-5 bullets]
- Tech stack: [list]
- Comp band: $[X] to $[Y] base, equity range [Z]
- Work arrangement: [Remote / Hybrid + city]
- The "why this role matters" pitch: [1-2 sentences]
Write a 350-450 word JD with sections: About the team, What you'll do,
What we're looking for, Tech stack, Compensation, How we hire.
Constraints:
- Lead with the team and the work (skip the company-first framing)
- Comp band is published (skip "competitive")
- Skip "fast-paced environment" and similar filler
- Direct second-person voice ("you'll own...")
Format as markdown ready to paste into Greenhouse or Lever.
12. De-bias a JD
You are an inclusive-hiring specialist auditing a job description for
gendered, ageist, and ableist language. Below is a JD draft.
[paste JD]
For each problematic phrase, give:
- The exact phrase
- Why it is biased (one sentence)
- A neutral rewrite
Format as a markdown table. If a phrase is acceptable, do not flag it.
Do not rewrite the entire JD, only the problematic phrases.
13. Rewrite a JD for senior-engineer audience
You are a senior technical writer who has hired 20+ staff engineers.
The JD below is written in HR-corporate language and will not attract
senior engineers. Rewrite it.
[paste JD]
Constraints:
- Lead with the actual engineering problem the role solves, in
concrete terms
- Specify the tech stack with versions and rationale
- Cut all "rockstar" / "ninja" / "passionate" language
- Cut required-experience years counts; replace with concrete output
examples ("you have shipped a system handling X")
- End with one paragraph on the engineering culture (specific
practices: how we do code review, how we ship, how we handle
on-call)
Use ONLY the facts present in the source JD. If the JD does not
specify stack versions, stack rationale, code review practices,
deploy cadence, or on-call structure, write the rewrite with a
placeholder like "[hiring manager: confirm exact version]". Do
not invent operational details. Flag every placeholder in a final
"TODO for hiring manager" section at the bottom.
Return the rewritten JD. Same length or shorter.
Screening prompts
Legal note first. Putting candidate names, resumes, or PII into ChatGPT raises real privacy concerns. OpenAI's default ChatGPT consumer tier may use inputs for training unless you opt out. Use ChatGPT Team / Enterprise (which excludes inputs from training by default) or Claude Enterprise for any prompt involving identified candidate data. We get into bias and compliance lower in the piece.
14. JD-to-resume fit scoring
You are a senior technical recruiter doing a quick fit pass on a
candidate resume against a job description.
Job description: [paste JD]
Candidate resume: [paste resume]
Generate:
- A 1-10 score on overall fit
- 3 specific strengths matching the JD requirements
- 3 specific gaps where the candidate does NOT match the JD
- One recommendation: "Move to phone screen," "Pass," or "Move to
technical screen"
Format as a structured response with each section labeled. Be honest
about gaps. Avoid being generous. Do not infer experience that is not on the
resume.
15. Red-flag detection on a senior-level resume
You are an experienced recruiter screening a senior engineering
candidate. The candidate is applying for a Staff Engineer role.
Resume below.
[paste resume]
Identify 5 potential red flags specific to senior-level hiring:
- Tenure issues (gaps, short stints at multiple roles)
- Title inflation (Senior or Staff title at companies where that
title is given out quickly)
- Stack mismatch with what we use
- Career-trajectory plateaus
- Missing senior-level signals (mentorship, system design, scope)
For each, cite the specific resume detail. Do not flag normal career
progression as a red flag.
16. Build a structured interview scorecard from a JD
You are a hiring manager designing a structured interview scorecard.
Below is the JD.
[paste JD]
Generate a scorecard for a 60-minute technical screen with:
- 4 evaluation criteria pulled directly from the JD
- 1-5 rating scale for each criterion with anchored descriptions
(what a 1 looks like, what a 3 looks like, what a 5 looks like)
- 2 interview questions per criterion designed to surface the signal
- A final hire/no-hire recommendation framework
Format as a markdown table. Anchored descriptions should be concrete
behavioral examples. Avoid adjectives.
Interview prompts
17. Behavioral interview questions tailored to the role
You are an interview-design specialist. Generate 8 behavioral
interview questions for a Senior Product Manager role at a B2B SaaS
company.
For each question:
- The question itself (open-ended, situation-based)
- The specific competency it surfaces
- A red-flag answer pattern (what a poor candidate sounds like)
- A green-flag answer pattern (what a strong candidate sounds like)
Avoid clichéd questions ("tell me about a time you failed"). Each
question should be one I would not regret asking 40 candidates in a
row.
Format as numbered entries with sub-bullets for each piece.
18. Technical-screen question generator (with calibration)
You are a senior engineer designing a 45-minute technical screen for
a Senior Backend Engineer hire. Our stack is Python + FastAPI +
PostgreSQL + Redis on GCP.
Generate 3 technical-screen questions:
- Question 1: a coding problem that takes ~15 minutes to write and
exercises Python data-structure intuition
- Question 2: a system-design question that takes ~20 minutes and
exercises database/caching trade-offs
- Question 3: a stack-specific question (Python, FastAPI, or
PostgreSQL internals) that takes ~10 minutes and exercises depth
For each, provide:
- The question prompt
- A "pass" answer outline
- A "strong" answer outline
- A common wrong-track to watch for
Format as 3 numbered sections.
19. Hiring-debrief synthesizer
You are facilitating a hiring debrief. Below are 4 interviewer
scorecards on the same candidate.
[paste 4 scorecards]
Synthesize:
- Where the interviewers agreed (consensus signals)
- Where they disagreed (calibration gaps to discuss)
- The overall recommendation based on the weighted signals
- One specific follow-up question to ask the candidate in a final
round if we proceed
Format as 4 labeled sections. Do not invent signals not in the
scorecards.
For the full breakdown of the senior-engineer sourcing process this connects into, see our senior software engineer sourcing playbook.
Hiring-manager intake prompts
20. Generate kickoff questions for a new role
You are a recruiter preparing a 30-minute intake meeting with a
hiring manager for a new role. The role is [Title] on the [Team] team.
Generate 15 intake questions organized into 4 sections:
- Role definition (5 questions: what success looks like at 30/60/90
days, what the role solves, etc.)
- Candidate profile (4 questions: must-have experience, must-have
skills, deal-breakers)
- Sourcing context (3 questions: where are these candidates today,
what are competitors paying, why have we struggled to fill this)
- Process logistics (3 questions: interview panel, decision timeline,
comp authority)
Format as a numbered list under each section header. Questions should
be specific enough that vague answers from the hiring manager will
expose gaps in their thinking.
21. Convert hiring-manager notes into a sourcing brief
You are a recruiter converting raw hiring-manager notes into a
structured sourcing brief.
Raw notes:
[paste unstructured hiring-manager notes]
Generate a 1-page sourcing brief with:
- Role title and team
- The 3 most critical 90-day deliverables
- Top 5 target companies (with one-line rationale for each)
- 3 disqualifying signals (deal-breakers)
- Comp band and work arrangement
- One contrarian sourcing angle the hiring manager mentioned that
might be high-signal
Pull EVERY specific (company name, deliverable, comp number,
disqualifier, contrarian angle) ONLY from the raw notes verbatim.
If the notes do not contain a top-5 target company list, output
"Target companies: [not specified in intake, propose in follow-up
meeting]" instead of inventing companies. Same rule for the
contrarian angle: if the notes do not contain one, write
"Contrarian angle: [none surfaced in notes]." Never substitute
your training-data knowledge for the hiring manager's expressed
preferences.
Format as a markdown 1-pager.
Where ChatGPT breaks (and what to use instead)
ChatGPT is excellent at the things above. It is also useless at four things you will hit by the end of week one.
No live candidate data. ChatGPT cannot pull current LinkedIn profiles, current GitHub activity, or live company headcount. It does not have a live web index for sourcing. The prompts above all assume you already have the candidate data; ChatGPT only helps you act on it.
No outreach automation. ChatGPT writes one message at a time. It does not send messages, track replies, drop responders out of sequence, or sync to your ATS. The 5-message sequence prompt above gives you the copy. You still need a separate tool to actually send the sequence at volume.
Hallucinated specifics. Stanford HAI found large language models hallucinate 58-88% of the time on specific legal queries (HAI, 2024). The same risk applies to recruiter use: ChatGPT will invent candidate names, fabricate contact details, and confidently misattribute career history if you ask it to fill gaps. Use ChatGPT to synthesize the data you bring. Avoid asking it to generate data you do not have.
Privacy and PII risk. OpenAI's consumer-tier ChatGPT may use your inputs for model training unless you opt out. Putting identifiable candidate data into a consumer-tier chat is a real compliance exposure. ChatGPT Team and Enterprise tiers exclude inputs from training by default; same for Claude Enterprise. Verify your tier before pasting any candidate PII.
No workflow integration. Every prompt above produces text. Getting that text into the tools your team actually runs hiring through requires copy-paste. At any meaningful hiring volume, this breaks.
This is where Yander fits. We built Yander because we hit these walls ourselves. ChatGPT handles the thinking: prompts, frameworks, templates. Yander handles the doing: sourcing across a 428M-profile index, automated multi-channel outreach with reply detection, and native integrations with the tools you already use (Slack, Notion, ClickUp, and more). Pricing is published: Free plan covers your first 200 candidates, Pro is $89 per month, Max is $249 per month, all on yander.ai/pricing. No placement fees, no setup fees, no contracts.
For the broader category map of where AI recruiting tools fit alongside ChatGPT, our AI recruiting software landscape for 2026 covers the full category. For LinkedIn Recruiter specifically, the LinkedIn Recruiter alternatives guide compares 12 tools side-by-side.
The legal landscape: bias, compliance, and what changed in 2025-2026
The compliance picture has moved sharply in the last 18 months. If you are using ChatGPT or any AI tool in hiring, this is the current state.
The Amazon case is the warning. Amazon scrapped its experimental AI recruiting tool in 2018 after internal testing found it penalized resumes containing the word "women's" and downgraded graduates of two all-women's colleges. The tool had been trained on a decade of historically male-dominated resume data and learned to replicate the bias (Reuters, Dastin, October 2018). Every AI hiring tool inherits the bias of its training data. ChatGPT is not exempt.
NYC Local Law 144 is in force. Effective January 1, 2023 with enforcement since July 5, 2023, NYC's Automated Employment Decision Tools law requires employers using AI in hiring decisions for NYC-based roles to: conduct an annual independent bias audit, post the audit results publicly, and notify candidates 10 business days before use. Penalties run $500 to $1,500 per violation per the NYC Department of Consumer and Worker Protection. A December 2025 audit by the NY State Comptroller found DCWP has struggled to enforce the law, with many employers neither disclosing AI use nor posting required audits.
Colorado AI Act is delayed but coming. Originally set for February 2026, then June 30, 2026, the Colorado AI Act has been pushed to January 1, 2027 after Governor Polis signed SB 26-189 on May 14, 2026. The law treats AI used in hiring, promotion, or termination decisions as high-risk and requires pre-use notice to candidates, an adverse-action process with human review, and 3-year record retention.
EU AI Act high-risk hiring obligations are proposed for deferral. AI in recruitment, candidate selection, and performance evaluation is classified as high-risk under Annex III of the EU AI Act. The high-risk obligations were originally set for August 2, 2026, but the AI Digital Omnibus proposes deferring them to December 2, 2027 (not yet final law). Fines can reach EUR 15M or 3% of global turnover for non-compliance with high-risk requirements, and EUR 35M or 7% for prohibited practices.
EEOC federal guidance is in flux. On January 27, 2025, the EEOC removed its May 2023 technical assistance on AI in hiring from its website following the new federal administration's executive order on AI policy. Federal-level guidance on AI hiring discrimination is currently unsettled. The underlying Title VII disparate-impact framework still applies; what specifically counts as compliant AI use is now unclear at the federal level.
Practical recruiter takeaway. If you operate in NYC, you need a bias audit before using AI in hiring decisions for NYC roles. If you operate in Colorado, prepare for January 2027 compliance now. If you have EU-based candidates, AI Act high-risk obligations are coming. Federally, the guidance is in transition, but the underlying disparate-impact law has not changed.
For startup-stage teams trying to design a hiring stack that stays compliant from day one, our best ATS for startups in 2026 guide covers the ATS-tier decisions that anchor a compliant hiring stack.
FAQs
Is it safe to put candidate names or resumes into ChatGPT?
It depends on the tier. OpenAI's consumer-tier ChatGPT (the free version and the $20/month ChatGPT Plus) may use your inputs to train future models unless you opt out in settings. That creates a privacy exposure for identifiable candidate data. ChatGPT Team ($25/user/month) and ChatGPT Enterprise both exclude inputs from training by default. Claude's Team and Enterprise tiers also exclude inputs from training. For any prompt involving identified candidate PII, use a Team or Enterprise tier. Avoid consumer tiers for PII. Read your tier's data-handling policy before assuming.
Can ChatGPT actually find candidates on LinkedIn?
No. ChatGPT cannot access LinkedIn profiles, GitHub repos, or live web data on its own. The "browsing" feature in some tiers can pull public web pages, but LinkedIn blocks ChatGPT's crawler. The prompts above for boolean search strings and X-ray queries help you write the search; you still need LinkedIn Recruiter, a sourcing tool like Yander, or a paid third-party index to run those searches against live candidate data.
What is the best ChatGPT model for recruiting work in 2026?
GPT-5 (released August 7, 2025) is OpenAI's current flagship and reports 45-80% lower hallucination rates than GPT-4o per OpenAI's published benchmarks. For long-context structured tasks like resume screening or scorecard synthesis, GPT-5 or Claude Sonnet 4.5 (released September 29, 2025, Anthropic) outperform earlier models on the published benchmarks from OpenAI and Anthropic. For quick drafts and rewrites, GPT-4o or Claude Sonnet 4 still work. The model that fits your team best depends on what your IT and compliance teams have approved for use with candidate data.
How do I stop ChatGPT outreach from sounding generic?
Generic outreach comes from generic prompts. The 5-part formula (Role, Context, Constraint, Format, Tone) at the top of this piece is the fix. Specifically: give the candidate signal in the prompt itself ("they shipped a React Native rewrite last quarter"), give the role specifics ("our payments team is 4 engineers, we ship daily"), and give a tone constraint ("respectful and direct, no flattering language"). ChatGPT mirrors the specificity you give it.
Can ChatGPT replace an ATS or sourcing tool?
No. ChatGPT writes text. An ATS stores candidate records, runs workflow, tracks pipeline stages, and integrates with your HRIS. A sourcing tool maintains a live candidate index, runs outreach at volume, and tracks replies. These are different jobs. ChatGPT pairs with both. It speeds up drafting inside whatever ATS and sourcing tool you already use.
What is the difference between ChatGPT and a dedicated AI recruiting tool?
ChatGPT is a general-purpose language model. A dedicated AI recruiting tool combines a model with a candidate index, outreach automation, and integration with your hiring stack. Yander, for example, runs sourcing across a 428M-profile index, sends multi-channel outreach you control, tracks replies, and connects to Slack, Notion, ClickUp, and more. ChatGPT helps you write better prompts to your AI tool. It does not replace the tool itself. Pricing on Yander is published: Free / $89 Pro / $249 Max on yander.ai/pricing.
Save these. Adapt the bracketed sections to your stack. When ChatGPT stops scaling for outreach volume, yander.ai/pricing is where the volume layer starts. Free for the first 200 candidates, $89/mo Pro after that.