👥 HR & Recruiting Prompt Generator Free · No Signup

Candidate Outreach Message Generator — Personalised, Without Sounding Automated

⚡ Quick Answer

A candidate outreach prompt works when it opens on a specific detail from the candidate's own work, names two or three genuinely specific things about the role, and asks for something small. This generator produces three message variants plus a follow-up bump in one run — each inside the platform character budget, none built on invented facts.

⚡ Candidate Outreach Message Generator — Personalised, Without Sounding Automated

Configure your facts below to generate a production-ready, constraint-locked prompt.

100% Client-Side

What this generator does

Cold outreach is where AI writing does the most damage in recruiting. Candidates can spot a templated InMail in two lines — "I came across your profile and was impressed…" — and the delete is instant. The problem is almost never the model's writing; it's the prompt. A prompt that says "write an outreach message for a backend engineer" gives the model nothing true to say, so it reaches for flattery and filler.

This generator is built around the two inputs that actually make outreach work. First, a specific detail from the candidate's real work — a project, a talk, a migration they described, a repo they maintain. The message opens on that detail, because it proves a human looked. Second, two or three genuinely specific selling points of your role — not "exciting opportunity at a fast-growing company", but "greenfield payments platform", "team of six, you'd be the first platform hire", "fully remote within India". Specifics are what a passive candidate weighs.

From those inputs you get three variants in one run — curiosity-led (opens on their work, teases the problem), outcome-led (leads with what the role achieves and why now), and peer-referral (framed as "your name came up in the context of…", for warm lists) — plus the follow-up bump, written at the same time, so your sequence is finished before you send message one.

Who it's for

How it works

  1. Find the detail first. Open the profile and pick one true, specific thing. If you can't find one, don't send outreach yet — a generic message to a named person is worse than none.
  2. Write selling points you'd defend in person. If you wouldn't say it in a call, it doesn't go in the message.
  3. Pick the channel. The generator formats to the budget: a connection-request note stays under 300 characters; an InMail gets the full structure.
  4. Generate all four messages at once. Three openers + the bump, all from the same facts.
  5. Send one variant, keep the bump dated. Follow up once, after 5–7 days, with the bump — one new reason, one yes/no question, under 60 words.

The prompt template

WORKED EXAMPLE / CODE 
You are an experienced recruiter writing candidate outreach that respects the reader.

INPUTS — use only these facts:
- Candidate first name: [name]
- Specific detail from their profile/project (must open the message): [detail]
- Their current role and company: [role, company]
- Role I am hiring for: [title, team]
- Genuine selling points of this role (2–3, specific): [list]
- Location / work model: [location, model]
- Sender: [your name, your title]
- Channel: [LinkedIn connection note / InMail / email]

WRITE FOUR MESSAGES:

MESSAGE 1 — CURIOSITY-LED: Open on the specific detail (what they built/did, in one line — no flattery adjectives). Name the problem or build this role owns. End with the low-friction ask.
MESSAGE 2 — OUTCOME-LED: Lead with what this role will have achieved in a year (from my selling points). Connect it to the direction their career already points. Same low-friction ask.
MESSAGE 3 — PEER-REFERRAL: Frame as "your work came up while we were mapping [problem area]". Shortest of the three. Same ask.
FOLLOW-UP BUMP: For 5–7 days later. Under 60 words. One genuinely new reason to reply (a new fact about the role, a deadline that's real, a related opening). One yes/no question. No guilt, no "just circling back", no fake "last attempt" unless it truly is.

FORMAT RULES:
- Connection note version of Message 1: maximum 300 characters including spaces.
- InMail versions: maximum 190 words each. Short paragraphs. No bullet lists in Messages 1–3.
- Address the candidate by first name once, naturally — not in every paragraph.

NEGATIVE CONSTRAINTS:
- NEVER open with "I came across your profile", "I hope this message finds you well", or a compliment about how "impressive" their background is.
- Do NOT invent facts about the candidate, their employer, or their work. If the specific detail is thin, the message must be shorter — not padded.
- Do NOT promise salary, title level, visa sponsorship, or timelines I did not supply.
- Do NOT describe the company with unverifiable superlatives ("world-leading", "revolutionary").
- Gender-neutral language throughout; no assumptions about the candidate's family, age, or circumstances ("now might be a good time to move because…" is banned).
Best model for this: Claude Sonnet 5.5 (natural, non-templated tone and tight instruction-following on length). GPT-6.1 works well when generating variants for a whole list from a spreadsheet of inputs.

Worked example

Filled inputs: Candidate: Priya. Specific detail: "Led the migration of her company's billing system from a monolith to services — wrote about the strangler-fig approach in a LinkedIn post." Current role: Staff Backend Engineer, fintech. Role: Principal Engineer, Platform. Selling points: "greenfield internal developer platform; team being built around this hire; remote within India". Sender: Arjun Mehta, Engineering Recruiter. Channel: InMail.

Output excerpt (Message 1, curiosity-led):

Hi Priya — your post on strangling the billing monolith caught my eye, specifically the call to keep the old invoicing path live until parity checks passed. We're hiring a Principal Engineer to build our internal developer platform from scratch — similar flavour of problem: replace the thing everyone depends on, without breaking it, except the "thing" is our deploy and tooling layer. The team's being formed around this role, and it's remote within India. Worth a 15-minute chat next week? — Arjun

Follow-up bump (excerpt):

Quick follow-up, Priya — since I wrote, the platform team's first mandate got confirmed: self-serve environments for 14 product teams. Still the same greenfield build. Open to a short call, or should I close the loop? (Under 60 words ✓, one new fact ✓, yes/no question ✓.)

Guardrails

Frequently Asked Questions

Practical answers covering model choice, workflow integration, and compliance rules.

Claude Sonnet 5.5 is the strongest choice for outreach specifically, because the quality that matters — a natural, non-templated voice that hits a word limit without sounding clipped — is where it leads. GPT-6.1 is the better pick when you're generating variants for a long list from structured inputs. What matters more than the model is the input: any frontier model writes convincing outreach from a real profile detail and real selling points, and none can rescue a prompt that gives it neither.

Work to the channel's budget. A connection-request note is capped at 300 characters — roughly two sentences: the specific detail and the ask. An InMail has room to work but shouldn't use it: around 150–190 words is the range where a passive candidate will actually finish reading. Follow-up bumps should be under 60 words. If your message needs more length, the problem is usually that it's selling the company instead of connecting the candidate's work to the role's problem.

Open on something only that candidate could receive — a named project, a specific decision they wrote about, a measurable thing they shipped. Generic personalisation ("your experience at [Company]") reads as a mail-merge field because it is one. Then keep the rest plain: what the role is, why it's interesting in concrete terms, one small ask. The generator enforces this by making the specific detail a required input and banning the stock openers ("I came across your profile…") that signal automation.

Once, after five to seven days — long enough not to feel like pressure, soon enough that the thread is warm. The bump must earn its place: one genuinely new piece of information (a confirmed mandate, a real deadline, a related role), one easy yes/no question, under 60 words. A second bump with no new information converts at a fraction of the first and costs goodwill; after the bump, close the sequence and let timing change the answer later.

Yes — the template is designed to be re-run per candidate with a table of inputs (name, detail, current role) against a fixed role block. Two cautions: the specific detail must be real for every row, so this only works if your sourcing notes carry one true detail per candidate; and bulk sending still has to respect platform rules and your own quality bar. Generate in bulk, review individually, send as a person.

Internal links

---