How to Write AI Prompts for Recruiting (With Copy and Paste Examples)

AI Recruiting Academy › Module 1: AI basics for recruiters

Why do short prompts give weak answers?

“Write a job description for a sales manager” leaves the AI to guess the industry, the level, the location, the pay, the team and the tone. It fills every gap with the most average answer it can produce, which is why the result reads like every other job ad online.

A language model predicts text from what you give it. Give it more of the right material and the prediction gets closer to what you actually need. Both OpenAI’s prompt engineering guide and Anthropic’s prompting overview make the same core point: clear, specific instructions with context and examples beat clever wording.

What are the five parts of a good recruiting prompt?

You will not need all five every time. For a quick rewrite, task and format are enough. For anything involving a candidate, always include the source material and the limits.

What does a weak prompt look like next to a strong one?

Six recruiting prompts you can copy

Replace the parts in square brackets with your own material. Remove names and contact details first if you are not on a business plan that your employer has approved.

1. Job description from intake notes

You are a recruiter writing a job ad for [company type]. Using only the intake notes below, write a job description with these sections: About the role, What you will do, What you need, Nice to have, Pay and benefits. Keep "What you need" to five points or fewer. Use plain, inclusive language and avoid phrases like "rockstar" or "ninja". If the notes do not include pay, write [PAY RANGE NEEDED] instead of guessing.

Intake notes:
[paste notes]

Then run the result through our free job description bias checker.

2. Boolean search string

Build a Boolean search string for [platform, e.g. LinkedIn Recruiter] to find [job title] candidates. Include common alternative titles and the key skills below. Only use operators that [platform] supports: AND, OR, NOT, quotation marks and parentheses. Explain each part of the string in one line.

Key skills: [list]
Titles to exclude: [list]

Test the string on the platform before trusting it, or build one with our Boolean search builder.

3. Resume against requirements

Compare the resume below with the requirements below. Make a table with columns: Requirement, Evidence from resume (quote it), Rating (Met, Partly met, Not shown). Only use what is written in the resume. Do not give an overall hire or reject recommendation.

Requirements:
[paste]

Resume:
[paste]

Notice the last line. The AI organizes evidence; you make the decision. Module 4 explains why that matters legally.

4. Personalized outreach

Write a first message to this candidate about the role below. Mention one specific, real detail from their profile that connects to the role. Under 80 words, conversational, no exclamation marks, no buzzwords. End with a low pressure question. Then write two alternative opening lines.

Role: [paste summary]
Profile: [paste]

5. Interview notes into a scorecard

Turn my interview notes below into a scorecard for the four competencies listed. For each competency, give two or three bullet points of evidence from the notes and flag anything I did not cover. Do not add anything that is not in my notes.

Competencies: [list]
Notes: [paste]

6. A kind, clear rejection

Write a short, respectful rejection email to a candidate who reached the final interview for [role]. Thank them for their time, give one honest line about the decision without legal risk (no reference to age, health, family or other protected traits), and invite them to stay in touch. Under 100 words.

How do you improve an answer you do not like?

Treat the first answer as a draft and reply to it, just as you would with a junior colleague.

Ask it to ask you first. Add “Before you start, ask me up to five questions about anything you need to know.” This is the fastest way to fix a vague brief.

Show an example. Paste an outreach message that got a great reply rate and say “match this style”. Examples steer tone better than adjectives.

Name the problem. “Too formal”, “cut it by half” or “the second paragraph repeats the first” works better than “make it better”.

Start fresh when you switch tasks. Long chats drift. A new chat with a clean prompt usually beats a tenth round of edits.

What should you never put in a prompt?

Keep out anything you would not email to a stranger: government ID numbers, bank details, health information, and full candidate files on a free personal account. Check your employer’s AI policy and the plan you are on. The privacy terms differ between free consumer apps and business plans, as Lesson 2 explains.

Quick answers

Do prompts work the same in ChatGPT, Claude, Gemini and Grok?

Mostly. The five part method works across all of them. Each model has quirks, so if you move a prompt between tools, test it on one or two real examples first.

Should I save my prompts?

Yes. Keep a shared document of prompts that work for your desk. Many assistants also let you save instructions as a custom assistant or project, so you do not retype the role and limits every time.

Is “prompt engineering” a skill recruiters need?

The basics, yes. SHRM’s State of AI in HR 2026 recommends training HR teams on prompt engineering alongside data literacy. You do not need to be technical; you need to be specific.

Why does the AI ignore part of my instructions?

Usually because the prompt is long and the key rule is buried. Put the most important limit at the end, repeat it once, and keep one task per prompt.