The recruiter prompt library: 30 prompts you can use today

A great prompt is just a clearly written brief. The best recruiters who use AI well treat their prompt library like recruiters used to treat their Boolean strings: build a small set of formats that work, refine them, share with the team.

Here are 30 to start with. Drop them into Claude, ChatGPT, or Vitae. Replace the [bracketed] bits with your specifics.

Sourcing & shortlisting

1. Write me three semantic search queries to find a [role title] in [location] with [must-have skill] and experience at [company type]. Make them different angles.

2. Read this candidate’s profile and score them 1-10 against these requirements: [paste rubric]. Show your reasoning per requirement.

3. Compare these three candidates against the role. Rank them. Tell me what each one is missing and what each one brings that the others do not.

4. Read these 20 candidate summaries. Group them into clusters by background pattern. Tell me which cluster is most likely to convert for a [role type] role.

5. Find me 5 second-degree connections of this candidate who fit the same role profile. Explain why each is worth reaching out to.

6. Generate a list of 20 companies whose engineering culture is similar to [target company]. Use signals like size, funding stage, tech stack, and team structure.

7. This candidate just changed jobs. Tell me whether that signal is positive or negative for a poach in 12 months and why.

8. Read this JD and write the ideal candidate profile in plain English. Two sentences max.

9. Re-rank my current shortlist if the role priorities change to: [new priority list]. Show the new top 5 and why.

10. From this list of 50 profiles, find me the 5 most likely to be open to a move based on tenure, recent promotions, and post history.

Outreach & follow-up

11. Read this candidate’s profile and give me three personalisation hooks I could use in an outreach message. Rank them by uniqueness.

12. Draft a 90-word outreach for [candidate name] for a [role title] at [company]. Lead with [hook]. Tone: warm, specific, no buzzwords.

13. Rewrite this outreach to be 30% shorter without losing the personalisation: [paste].

14. Suggest five subject lines for this outreach. Format each one as ‘Name - specific thing about them’.

15. Draft a follow-up message for someone who opened the email twice but did not reply. Tone: light, no pressure.

16. The candidate replied with this objection: [paste]. Draft a thoughtful response that acknowledges the concern and offers a small next step.

17. Write a re-engagement message for a candidate I reached out to 6 months ago. Reference what has changed since.

18. Compare these two outreach drafts. Which is more likely to get a reply and why?

19. Draft a polite decline for a candidate who is a strong human but not a fit for this specific role. Keep the door open.

20. Write a referral ask for a placed candidate, three months in. Tone: warm, specific to their experience.

Screening & interviews

21. Generate five technical screening questions for a [role title]. For each, tell me what a strong, average, and weak answer looks like.

22. Summarise this 45-minute interview transcript into: motivation, comp expectations, signal strengths, risk flags, recommended next step.

23. Score this candidate against the role rubric based on the interview transcript. Cite the moments that informed each score.

24. Generate three behavioural questions to probe whether this candidate has actually led a team versus just managed people.

25. Compare this interview transcript to my last three transcripts for the same role. What patterns am I missing?

JDs, market research, reporting

26. Rewrite this JD to be 40% shorter. Lead with the impact of the role, not a list of bullet points. Remove gendered language.

27. Benchmark salary for a [role title] in [location] with [years] of experience. Cite three sources and explain the range.

28. Write a market map of the top 15 companies hiring [role type] in [location] right now. Include the hiring manager if public.

29. Summarise this week’s pipeline for me. Flag stalled candidates, candidates I have not contacted in over 7 days, and any risk flags.

30. Convert this messy CSV of candidate notes into a structured table with columns: name, current role, last contact, status, next action.

How to get more out of these

• Always give the AI context first. The role, the rubric, the company. Skip this and you get generic output.
• Treat the first answer as a draft. Push back, ask for variations, ask why.
• Save the ones that work. Build a personal prompt library you can pull from.
• Share with your team. The best prompt libraries get refined by ten people, not one.