ChatGPT for Recruiters: 15 Practical Prompts, Examples and Clear Limits

Why ChatGPT for recruiters is handy once you set tight limits

Recruiters often use AI for time-consuming tasks that look very similar in content. Think of writing a job advert with ChatGPT, setting up a first search direction, coming up with ChatGPT intake questions, and testing AI message variants. This is enormously useful, since it gets you from a blank screen to a usable first draft much faster. Even so, keeping an eye on limits still matters. After all, a language model doesn't automatically know your vacancy, your team, your pay policy or a candidate's context. That's why ChatGPT for recruiters only works well if you clearly state what's already fixed and what the model specifically must not fill in itself.

When using ChatGPT in recruitment, the benefit mainly lies in speed and structure. The model helps you phrase, rewrite and organise. It is not, however, a source of truth, and it should never take decisions off your hands. Anyone who keeps that distinction clearly in mind gets to strong drafts faster and keeps a firm grip on quality themselves.

How to use ChatGPT for recruiters with a fixed prompt structure

The four building blocks of a good prompt

Writing a good prompt starts with four fixed parts: role, context, example and the desired format. With the role, you decide which hat the model wears, for instance recruiter, sourcer or editor. With the context, you share only the facts that are actually known. Through an example, you show the tone, structure or writing style you have in mind. Finally, with the format, you ask for a clear end result, such as three separate variants, a short overview or an adjusted search string. Using this more often within a team? Then a Custom GPT for recruiters is often the handiest choice, because it lets you safeguard fixed instructions and the required context far more reliably.

A quick self-check for every prompt

Run exactly the same check for every prompt. First look at which facts you've actually supplied, then decide which assumptions are absolutely off limits. Finally, check what you still need to verify yourself. This simple step stops a neatly written text from coming across as trustworthy when it shouldn't. That risk is real with examples of recruitment prompts, because a language model often phrases its answers very convincingly.

When fixed instructions work better than writing prompts from scratch each time

Individual prompts work fine for personal use. For teams, though, fixed instructions quickly become the more logical choice; otherwise tone, quality and structure can vary quite a lot per recruiter. You notice this most with job adverts, sourcing and messages to candidates. When everyone works to the same rules, the output is far more consistent and checking it takes considerably less time.

3 prompts for ChatGPT for recruiters on job adverts

Prompt 1: rewrite a job advert to B1 level

Goal: make an existing job advert simpler and clearer, so more candidates quickly understand the content.

Prompt: You are a recruitment copywriter. Rewrite the job advert below to B1 level, in clear English. Keep the facts unchanged. Don't invent extra terms of employment, culture claims or duties. Use short sentences and concrete words. Structure the output in this order: short intro, duties, must-haves, nice-to-haves, what's on offer, application step. Job advert: [paste anonymised job advert].

Pitfall: the model may fill in benefits, a certain atmosphere or terms of employment that aren't in the source text.

Self-check: check carefully that the salary, the number of hours, the team information and the terms exactly match the source.

Prompt 2: improve inclusive wording

Goal: make a job advert more neutral and accessible, without changing the substance.

Prompt: You are an editor for inclusive job adverts. Analyse the text below and rewrite sentences that unnecessarily exclude people, are vague, or seem too targeted at one specific type of candidate. Keep the content and the seniority level unchanged. First give five points of attention, then share an improved version. Use clear, plain English. Text: [paste anonymised job advert].

Pitfall: the model can make the text too generic, causing the job description to lose its sharpness.

Self-check: check that the hard requirements still come across clearly and that the seniority level has stayed the same.

Prompt 3: separate requirements from preferences

Goal: create a clear distinction between what's genuinely essential and what's merely desirable.

Prompt: You're helping a recruiter sharpen a vacancy. Read the text below and split all the requirements into two groups: must-haves and nice-to-haves. Give a short, understandable reason for each point. Then add three questions I can ask the hiring manager if the list is still too broad. Use only the information in this text: [paste anonymised job advert].

Pitfall: the model may pick its own priorities, while these haven't officially been decided internally yet.

Self-check: always check the outcome against the intake, and get the hiring manager to explicitly confirm what's genuinely a requirement.

What to watch for when writing a job advert with ChatGPT

When writing a job advert with ChatGPT, one simple rule of thumb applies: never let the model fill in facts you haven't supplied yourself. In practice this often goes wrong with the description of company culture, growth opportunities, terms of employment and team composition. So always use a reliable source, such as a thorough intake or an existing text. Working with fixed formats and a specific tone of voice? Then templates and instructions for recruiters help keep texts considerably clearer and more consistent across the team.

3 prompts for ChatGPT for recruiters on a Boolean search

Prompt 4: set up a search strategy around a job title and synonyms

Goal: quickly build a first list of job titles, variants and useful synonyms for sourcing.

Prompt: You are a sourcer. For the role [e.g. backend developer], build a search strategy in English. Give job titles, common variants, synonyms and related roles. Set out the output clearly in four blocks: core job titles, alternative titles, technical terms and related roles. Don't use any candidate names, and don't make assumptions about seniority unless it's explicitly stated.

Pitfall: the list can quickly become too broad, so your search returns too much noise.

Self-check: cut the titles that are simply too far removed from the specific role you're searching for.

Prompt 5: generate exclusions and alternative title variants

Goal: create a search setup with terms you specifically want to include as well as terms you want to explicitly exclude.

Prompt: You're helping with a Boolean search via ChatGPT. For this role: [role] and target group: [short description], make a list of inclusions and exclusions. Give ten relevant job titles, ten alternative spellings and eight possible exclusions that can reduce noise. Also explain, per exclusion, in just one sentence why it's specifically useful.

Pitfall: exclusions can sometimes be phrased too sharply, wrongly ruling out suitable candidates.

Self-check: check extra carefully that you're not excluding groups in a way that's unfair or even discriminatory.

Prompt 6: sharpen the search string when the result is too broad or too narrow

Goal: improve an existing search string based on too many or too few candidates found.

Prompt: You are a sourcing specialist. I'll give you a search string and the specific problem I'm having. Analyse the string and give me three improved variants: a narrower version, a broader version and a balanced version. Explain, per variant, in plain language what the likely effect will be. String: [paste string]. Problem: [too broad or too narrow, with a short explanation].

Pitfall: the explanation can sound logical but still not fit well with the underlying search logic of the platform you're working in.

Self-check: test every string in the actual channel and pay close attention to brackets, quotation marks, active filters and unexpected noise in the results.

What a Boolean search with ChatGPT does and doesn't do

ChatGPT is a great help for thinking through synonyms, search directions and alternative job titles. The model, however, doesn't automatically know how your specific market, channel and target group are behaving right now. So use it mainly as preparation for searching for candidates with ChatGPT, and absolutely not as a final check. The recruiter always stays responsible for the final search string and for setting fair, workable filters.

3 prompts for ChatGPT for recruiters on target group research

Prompt 7: map out job variants and adjacent roles

Goal: spot even faster which job titles and role variants sit in the same space.

Prompt: You're carrying out exploratory target group research for recruitment. For the role [role], I want a handy overview of adjacent roles, job titles and specialisations. Split the output into three blocks: direct variants, adjacent roles and possible progression paths. Keep the output factual, and don't state anywhere that this is current market data.

Pitfall: the model may select related roles a little too broadly.

Self-check: check that the roles listed genuinely fit your specific vacancy and the required seniority level.

Prompt 8: explore adjacent sectors

Goal: discover new pools for sourcing, especially when the primary target group is fairly small.

Prompt: You're helping a recruiter effectively explore alternative target groups. For this role: [role], and the current sector: [sector], give me five adjacent sectors where similar work experience could be useful. Also explain, per sector, which overlap in work, knowledge or working environment is relevant. Explicitly label this as assumptions, and don't present anything as a hard labour-market fact.

Pitfall: the model may name sectors where a move in practice has very little chance of succeeding.

Self-check: test the outcome at a later stage against real market information or concrete results from your sourcing.

Prompt 9: draft assumptions for later research

Goal: formulate clear research questions you can later verify with reliable data.

Prompt: You're supporting labour-market research via ChatGPT. For the role [role], formulate eight assumptions I can later verify with real sources. Spread them across job title, location, experience, sector background, barrier to switching, salary expectations, competition and messaging. Keep each assumption short and neutral. Then add one check question to each assumption.

Pitfall: the output sometimes sounds very convincing, while it's still purely a set of working hypotheses.

Self-check: explicitly mark every outcome as a working version and only consult reliable sources afterwards for confirmation.

What to watch for with labour-market research via ChatGPT

A language model is never an up-to-date labour-market source. That's an important limit within ChatGPT for recruiters, since choices around target groups quickly become quite strategic. So use the model mainly for coming up with questions, fresh angles and useful prompts for recruiters. Rely on real data sources, though, when it comes to numbers, competition, regional analysis and salary ranges.

3 prompts for ChatGPT for recruiters on intake preparation

Prompt 10: intake questions about results and context

Goal: sharpen the intake call considerably with questions about the actual output of the role.

Prompt: You're helping a recruiter with intake preparation. Based on this job description: [text], create twelve intake questions that specifically address results, team context, priorities, collaboration, pace and the expected success in the first six months. Use plain, everyday language and avoid generic questions that don't deliver much in practice.

Pitfall: the questions sound good at first, but sometimes turn out too generic for good, substantive decision-making.

Self-check: select only the questions that genuinely sharpen the sourcing, the selection process or the final job advert.

Prompt 11: sharpen must-haves and rejection criteria

Goal: get crystal clear together with the hiring manager on what's genuinely a hard requirement.

Prompt: You're supporting a recruiter. I'll give you a rough intake for a vacancy. Turn it into three clear lists: must-haves, negotiable points and hard rejection criteria. Then add six deeper ChatGPT intake questions to quickly resolve any points of doubt. Use only the information given. Intake: [paste anonymised intake].

Pitfall: the model may phrase the rejection criteria far too strictly, while internally there's actually still some flexibility.

Self-check: have the hiring manager explicitly confirm each possible rejection criterion before you actually select on it.

Prompt 12: surface missing information

Goal: see at a glance what's still missing before you actually open a vacancy or start sourcing.

Prompt: You are a recruiter and quality controller at the same time. Analyse this intake and identify which information is currently still missing to be able to search, write and assess effectively. Split the output clearly into four blocks: role content, team and context, hard requirements, and terms of employment and process. Also add three concrete follow-up questions per block. Intake: [paste anonymised intake].

Pitfall: in its phrasing, the model quietly fills in some of the missing facts anyway.

Self-check: check critically whether every question comes from an actual gap in the intake, and not simply from an assumption made by the AI model.

How to use ChatGPT intake questions without inventing facts

AI can certainly help with your intake preparation to add structure and sharpness. The model, however, should never simply present missing vacancy facts as absolute certainty. This efficient way of working fits agencies especially well, since intake, sourcing and candidate contact follow each other there at high speed. In that busy setting, using ChatGPT for recruitment agencies helps bring overall speed and quality far more closely into line.

3 prompts for message variants with AI

Prompt 13: tone variants for a first contact

Goal: create multiple tone angles for a first approach, without using any personal data.

Prompt: You write first-ever recruitment messages. Create three variants for the same role: businesslike, warm and nicely direct. Use this context: the role [role], the reason for contact [short reason], the proposition [factual core] and the call to action [desired step]. Keep each message concise, under ninety words, and don't invent any personal details yourself.

Pitfall: the messages are linguistically tidy, but stay a little generic in content.

Self-check: check whether the trigger, the specific role and the proposition have genuinely been worked out concretely enough.

Prompt 14: test opening angles

Goal: compare different openings properly against each other for exactly the same message.

Prompt: You're helping to come up with AI message variants. Write five strong openings for a first approach message about this vacancy: [short context]. Use five different angles here: the content of the role, the impact to be made, the team context, perfect timing, and learning opportunities. Also write a very short explanation, per opening, of when that version fits best.

Pitfall: openings sometimes sound very clever, but in practice still don't fully match the reality of the role.

Self-check: check whether you can trace every opening straight back to the hard facts from the vacancy or the original intake.

Prompt 15: vary the call to action and the length

Goal: cleverly test the ideal message length and closing line for different communication channels.

Prompt: You are a recruiter. Rewrite this recruitment message in three specific lengths: fifty words, eighty words and a hundred and twenty words. Also give each version a slightly different call to action: one low-effort, one concrete and one strongly geared towards booking a call. Keep all the facts, don't invent a personal trigger yourself, and keep the tone professional at all times. Message: [paste anonymised text].

Pitfall: very short versions quickly lose valuable nuance or credibility.

Self-check: read every version created out loud and check whether the message still sounds human and completely clear.

Why AI message variants aren't yet personal outreach at scale

A generic chatbot is extremely strong at endlessly varying and rewriting. That doesn't automatically mean, though, that you instantly have successful, personal outreach at scale in your hands. For that, you need context, a critical review, a consistent tone of voice and a tight working process. Especially with AI prompts for hiring, it simply holds true that a first draft only really adds value once it fits perfectly with the role, the intended target group and the chosen channel.

These 5 tasks you don't hand over to AI in recruitment

Assessing candidates

An AI model can absolutely help you summarise information or prepare good questions. It should, however, never take over an actual candidate assessment. Assessing someone always calls for human judgement, rich context and taking responsibility. The risk of unfair or bias-confirming assumptions is simply too great.

Final rejection

A rejection almost always affects someone directly. That's why the final decision and the final check should always sit with a real human. AI can, at most, draft a rough message for you in this process, which you then fully shape yourself, both in content and in tone.

Deciding on a salary

Setting a salary brings major business, legal and relational consequences with it. A chatbot never fully knows your own reward policy, sensitive internal dynamics and the exact room for negotiation. Use AI here only to prepare your questions or for a quick summary of information you already have.

Entering personal data without proper agreements

Never simply enter candidate data into a tool without clear internal agreements in place for this. The single safest rule? Use as little data as possible. Preferably work with fictional or carefully anonymised examples when testing new prompts. This is always easier to explain internally and significantly reduces unnecessary risk.

Sending messages without checking them

Even good AI output can quietly still contain stubborn mistakes. That's why, as a recruiter, you should never send these messages blindly. Always weave a short, fixed check for factual accuracy, tone, clarity and appropriate language into your daily working process.

Privacy and control with ChatGPT in recruitment

Personal ChatGPT environments and model improvements

Within personal ChatGPT environments, data you enter is used by default to further improve the model, unless you actively turn this off as a user. That small nuance matters enormously for busy recruiters who just want to quickly copy and paste a bit of text. So never put personal data, loose CV sections or other sensitive information into your prompt if you don't have a legal or internal basis for it.

Business AI products and your organisation's data

When using paid, business products, according to OpenAI's own help information, shared organisation data is by default no longer used to train new models. Note: that doesn't take away your own professional responsibility. Clear internal permission, strict privacy rules and deliberate data minimisation remain essential, simply because the recruitment sector is packed with sensitive information and business context.

A practical rule for recruiters

When a generic chatbot is no longer enough

Individual prompts work brilliantly for thinking and writing work

Individual prompts work great for a quick first draft, a few handy variants or concise summaries. That applies to ChatGPT prompts for recruiters and countless other everyday writing tasks. You reach the limit, though, as soon as multiple team members start working together on the same vacancies, intake processes and outreach campaigns. From that point on, structural repeatability becomes far more important than pure, individual speed.

Why dedicated recruitment software works better for context, reviews and process

Precisely from that tipping point, rich context becomes more valuable than loose bits of AI output. In that kind of situation, after all, you want fixed instructions, built-in review steps, smart reusable formats and full control over your entire process. You see this pattern nicely reflected, for example, in the Manpower case on personalised InMails. There, a smart workflow was deliberately put in place, packed with tight templates, a review layer and processing in efficient batches. In the end, this project measured a solid 43 percent response rate as a client result. Of course, a percentage that high is no cast-iron guarantee for other recruitment teams, but it does perfectly underline why a well-embedded way of working is often many times stronger than loose text generation through a simple prompt.

For teams: put fixed instructions and a consistent tone of voice in place

Virtually every recruitment team benefits enormously from well repeatable instructions. Without that framework, your quality depends far too much on whichever colleague happens to write the prompt. Tight, fixed frameworks help everyone hold onto exactly the same tone, structure and level of quality control in job adverts, intakes and the final candidate messages. This makes the necessary review step run far faster, and it directly reduces the chance of annoying, inconsistent style differences between recruiters.

For agencies: where extreme speed and high quality need to come together

For agencies, there's something extra going on in the background. Tasks such as intake, sourcing and the final candidate contact often all run at the same time here. That high speed is crucial for the business, but so, of course, is your reputation as an agency. A generic chatbot might get you out of a bind quickly for a bit of loose text. An integrated recruitment workflow, however, works many times better if you genuinely want to bring durable quality, rich context and clear team agreements together structurally.

Try it yourself with a completely safe example

Pick any one prompt from this article and try it out with a fully anonymised example. So make sure you absolutely don't use any real, current candidate data. Then take a very critical look at the AI output. Is the chosen tone right, are the facts included fully complete, and has the model itself slipped in any noticeable assumptions anywhere? If you want to experience this important difference for yourself, you can use the page try a recruitment message yourself to run a safe example and then critically compare the AI output with your own trusted way of working.

Frequently asked questions

Is ChatGPT suitable for recruiters who have very little time?

Yes, mainly for recurring writing work and repetitive thinking work. Think of job adverts, thorough intake preparation, setting up a Boolean search and generating snappy message variants. The real time savings, though, only really kick in once you consistently work with fixed prompt structures and add a quick, structural check to your process.

Am I allowed to put information from a CV into a prompt?

It's best not to do this without thinking, unless you have clear internal agreements about it and a valid legal basis for it. When testing, preferably always work with carefully anonymised or entirely fictional examples. This approach fits far better with applicable privacy rules and lowers the risk of unnecessary and risky data sharing.

What's the best way to get started with prompts for recruiters?

Simply start with one fixed task that comes up often every week, such as quickly rewriting a job advert or preparing sharp intake questions. Consistently use the same structure each time, with a role, the context, a good example and a tight desired format. Only then critically measure whether the new output genuinely saves you time now, and whether it actually calls for fewer corrections.

Can ChatGPT independently select or reject candidates?

No. AI can absolutely help you summarise data, structure processes and prepare questions. The real selection and the final rejection, however, remain human work, simply because human responsibility, a lot of context and a fair judgement are inseparable from that.

When do you successfully move from individual prompts to a fixed workflow?

That tipping point usually arrives once multiple recruiters are actively working on the same processes internally, or once the final quality turns out to vary too much per individual user. From that point on, fixed AI instructions, a review process and strict process control matter far more than pure, individual speed. That's exactly why growing teams often choose a well thought-through approach in which the required context and the shared way of working are far better embedded in the system.

In short: ChatGPT supports the thinking and writing work fantastically, but the recruiter always remains personally responsible for the accurate facts, the right tone and the final decision.