ChatGPT prompts for recruiters: 10 copy-ready prompts with checks

The ChatGPT prompts that work best for recruiters give the model a role, the source material to work from, the exact output format, and a rule against inventing facts. Below are ten copy-ready prompts for intake calls, job ads, sourcing, outreach, phone screens, ATS notes, rejections and client updates, each with what to paste in and what good output looks like. Before you paste anything, check what your plan does with your data: as of October 2026, OpenAI says it may use content from Free, Plus and Pro accounts to train its models unless you opt out, and does not by default for its Business, Enterprise and API services.

Prompts for client submittals have their own page, ChatGPT prompts for candidate submittals, and prompts for building interview questions and scoring guides are in AI prompts for interview questions. The prompts work in other assistants too; only the data terms differ.

What can you safely paste into ChatGPT?

OpenAI's help article on how your data is used to improve model performance says, as of October 2026:

• Content from individual plans (Free, Plus and Pro) may be used to train its models. You can opt out by turning off Improve the model for everyone under Settings, then Data controls, or through its privacy portal.
• Even after opting out, giving a thumbs up or down on a response may let OpenAI use that whole conversation for training.
• Temporary chats are not used to improve its models while they remain temporary.
• By default, inputs and outputs from ChatGPT Business, ChatGPT Enterprise, ChatGPT Edu and the API are not used to improve its models.

A training opt-out is not the same as permission. Whether you may paste a candidate's resume or your interview notes into any tool depends on your organization's policy, your client agreements, what the candidate was told, and privacy law. A workable default:

The policy behind this, including the risk tiers and the AI register, is in how to use AI in recruiting responsibly.

What makes a recruiting prompt work?

Every prompt below has the same five parts, and it is worth keeping them when you write your own:

1. Role: who the model is writing as, such as an agency recruiter or a hiring manager.
2. Source: the job description, your notes or the facts it may use, pasted in full.
3. Task and format: exactly what to produce, with headings, length and tone.
4. Rules: what to avoid, including a ban on facts that are not in the source.
5. A gap marker: an instruction to write [MISSING] instead of guessing.

Which prompts help with intake and job ads?

1. Prepare questions for an intake call

You are an experienced recruiter preparing for an intake call with a
hiring manager. Below is the job description for [JOB TITLE].

Write 12 questions for the call, grouped under: the problem this hire
solves, must-have skills, deal-breakers, what success looks like at
90 days, the interview process, and pay and logistics.
For each must-have in the job description, add one question that tests
whether it is truly required.
Use only the job description. Where something is unclear, turn it into
a question rather than assuming an answer.

Job description:
[PASTE JOB DESCRIPTION]

• Paste in: the job description, with the client's name removed if the search is confidential.
• Good output: questions specific to this job ("Is warehouse system experience required, or would general ERP experience transfer?") rather than generic ones. Compare with the full list in intake meeting questions.
• Check facts and data: delete any question that assumes a fact the description does not state. Confidential client details follow the data rules.

2. Turn a job description into a job ad

Write a job advertisement for [JOB TITLE] using only the facts in the
job description and notes below.
Structure: a two-sentence opening about the work itself; "What you'll
do" (5 bullets); "What you need" (must-haves only, maximum 5); "Nice to
have" (maximum 3); pay range and location exactly as given; how to apply.
Plain language. No "rockstar", "ninja", "fast-paced" or "family". No
requirements about age, years since graduation or physical ability
unless they appear below.
If pay, location or working pattern is missing, write [MISSING: item].

Job description: [PASTE]
My intake notes: [PASTE, NO CANDIDATE DATA]

• Paste in: the description and your intake notes.
• Good output: an ad a candidate could self-screen against, with a short must-have list. How to write a job description that screens explains why short beats long.
• Check facts and data: every requirement and benefit against the source, and the pay range against the rules where the job is advertised; see pay transparency laws by state. No candidate data is needed here; see the data rules.

3. Separate must-haves from nice-to-haves

Here is a job description and my intake notes. Put every requirement
in a table with columns: requirement | must-have or nice-to-have |
evidence from the notes for that choice | one screening question that
would test it.
Flag any requirement that might screen out capable people without being
needed for the work (for example a degree for skills the notes say are
learned on the job), with a one-sentence reason.
Use only the text below.

[PASTE JOB DESCRIPTION]
[PASTE INTAKE NOTES]

• Paste in: both documents.
• Good output: three to five must-haves, each with a reason you could repeat to the hiring manager.
• Check facts and data: the hiring manager, not the model, decides what is required. Use the table as the agenda for that conversation, and keep client-confidential notes within the data rules.

Which prompts help with sourcing and outreach?

4. Build Boolean search strings

Build Boolean search strings for [LinkedIn Recruiter / Google X-ray /
our ATS] to find [JOB TITLE] candidates.
Must-haves: [LIST]. Locations: [LIST]. Exclude: [e.g. recruiters, students].
Give me: (1) a broad string, (2) a narrow string, and (3) the alternative
job titles and skill synonyms you used, as a list I can edit.
Use only operators the platform supports. If you are unsure whether an
operator is supported, say so.

• Paste in: the must-haves from prompt 3, not the whole description.
• Good output: balanced parentheses, synonyms people actually use on profiles, and a note on any operator it was unsure about.
• Check facts and data: run both strings and read the first 20 results before trusting either. Method and platform limits are in how to source candidates on LinkedIn. The prompt needs no personal data at all.

5. Write a first outreach message

Write a first outreach message to a passive candidate for [JOB TITLE].
What their public profile shows: [2-3 FACTS, e.g. "six years in
warehouse operations; opened a new site in 2024"].
The role: [3 FACTS, INCLUDING THE PAY RANGE].
Rules: under 90 words; one specific point from their profile and why it
fits; end with one clear question; no flattery or "impressive profile";
claim nothing about the role that is not listed above.
Give two versions: LinkedIn InMail, and email with a subject line.

• Paste in: two or three facts from the public profile, written by you. No contact details.
• Good output: a message that could only have been sent to this person. More structures are in candidate outreach email templates.
• Check facts and data: the profile facts are right and the pay range matches the job. Never paste a candidate's contact details or anything beyond their public profile (data rules).

6. Write a follow-up sequence

Here is my first outreach message: [PASTE].
Write two follow-ups for when there is no reply: one after 4 days that
adds one new fact about the role, and one after 10 days that closes the
loop politely and asks whether they know someone suitable. Each under
60 words, same tone, no repetition of the first message.
New fact I can use: [ONE FACT ABOUT THE ROLE OR TEAM]

• Paste in: your sent message and one new fact.
• Good output: short messages that give the candidate a new reason to reply, not "just bumping this."
• Check facts and data: the new fact is true and approved for sharing by the client, and the thread contains no candidate details you may not share (data rules).

Which prompts help with screening?

7. Draft phone screen questions

Using the job description and must-haves below, write a 20-minute phone
screen: first, logistics questions (location, notice period, pay
expectations, and whether they are authorized to work in [COUNTRY],
asked the same way for every candidate); then one question per must-have,
each with a follow-up probe and one line describing a strong answer.
Do not include questions about age, family, health, religion,
nationality, or salary history.

[PASTE JOB DESCRIPTION AND MUST-HAVES]

• Paste in: the job description and your must-have list.
• Good output: questions you could ask word for word, with probes. Compare with phone screen questions for recruiters.
• Check facts and data: remove anything unlawful where you hire, including salary history questions in states that ban them; see salary history ban states. The prompt needs no candidate data (data rules).

8. Turn rough screening notes into an ATS note

Turn my rough screening notes into a structured ATS note with these
headings: Summary (2 lines); Must-haves (met / not met / unclear, with
the candidate's words where I noted them); Motivation; Pay and notice;
Concerns; Next step.
Rules: use only what is in my notes. Do not add skills, numbers or
opinions. Keep quotes exactly as written. If a heading has nothing,
write "Not covered."

Notes (name replaced with initials):
[PASTE NOTES]

• Paste in: your own notes, with the name replaced by initials, and only in an account approved for candidate data.
• Good output: a note that says less than your rough notes, not more. The structure follows how to write candidate notes in your ATS.
• Check facts and data: every quote and number against your notes, and strip anything about health, family or other protected traits before saving. If your account is not approved for candidate data, do not use this prompt (data rules).

Which prompts help with candidate and client updates?

9. Write a rejection email

Write a short rejection email to a candidate interviewed for [JOB TITLE]
who was not selected. Warm, direct and final.
Include: thanks for their time; the decision within the first two
sentences; this specific positive from my notes: [ONE GENUINE POINT];
whether we will keep their details for future roles: [YES/NO].
Do not give reasons for the decision, do not invent feedback, and do not
promise future contact unless I said yes. Under 120 words.

• Paste in: the role title and one real positive you noted. No other candidate data is needed.
• Good output: a message a candidate can read once and understand. Other versions are in candidate rejection email templates.
• Check facts and data: the positive point is true, and the email does not hint at a reason you would not want to defend. Leave out the candidate's name and other details until you paste the result into your email (data rules).

10. Write a weekly client update

Write a weekly update email to a hiring manager about our search for
[JOB TITLE].
This week's facts: [profiles reviewed, people approached, screens done,
submittals, interviews booked]. Market feedback: [e.g. "three candidates
declined at the stated pay range"]. What I need from the client: [e.g.
"feedback on two submittals by Thursday"].
Structure: status in one line; the numbers as a short list; one market
insight and what it means; one clear request with a date. Under 150
words. Use only the numbers I gave you.

• Paste in: your own activity numbers and feedback, with candidates referred to as A, B and C.
• Good output: an update a busy manager can act on from a phone.
• Check facts and data: every number matches your ATS, and no candidate is named in the prompt (data rules).

Which prompts should recruiters not use?

• "Rank these candidates" or "Who should we hire?" OpenAI's usage policies restrict automating high-stakes decisions in areas including employment without human review, and evaluation tools carry legal duties in some places; see AI hiring laws by state.
• Anything that infers protected traits, such as guessing age, ethnicity or nationality from a name, photo or graduation year.
• Finding personal contact details or private accounts for a named person.
• Writing feedback or references you have not formed yourself. If the opinion is not yours, it should not go out under your name.

How do you check ChatGPT's output before using it?

1. Compare every name, number, date, employer and pay figure with the source.
2. Search the output for [MISSING] and fill each gap yourself or delete the sentence.
3. Read it aloud once. If it does not sound like you, edit it until it does.
4. Remove anything about protected traits or details the candidate did not agree to share.
5. Save the final version, not the draft, wherever your records live.