Marcus Bell sources senior infrastructure engineers for a 220-person fintech company, working a req list that is almost entirely passive: the people he wants are employed, well paid, and not reading job boards. In a normal week he sends roughly 40 LinkedIn Recruiter messages and gets back 4 or 5 replies, and most of those are polite declines. He suspects the problem is not his roles but his messages, which read like everyone else's. This lesson follows Marcus as he rebuilds his passive outreach around real research and AI-assisted drafting, the kind that earns a reply because the candidate can tell a human actually looked at their work.

Why Passive Engagement Is a Different Job

An active candidate has already decided to look. They expect recruiter messages, they want to hear about the role, and a clear job description with a salary band can be enough to start a conversation. A passive candidate like the ones on Marcus's list has decided nothing. They are not unhappy, they are not browsing, and they delete most recruiter outreach within a few seconds. The bar is not "is this a good job," it is "is this worth interrupting a week I am already content with." Passive candidates are simultaneously the hardest people in recruiting to reach and the most valuable to land, and those two facts are related: they are hard to reach precisely because they are good enough that everyone else is writing to them too.

That changes the goal of the first message. Marcus is not trying to sell the role. He is trying to earn a single reply by proving he understands who this person is and why this specific opportunity might matter to them. The job description is almost irrelevant in the opening touch. What matters is evidence of attention. A passive candidate who feels genuinely seen will reply even to say "not now, but stay in touch," and that reply is the entire win of the first message. The messages passive candidates do respond to all share one property: they demonstrate that the sender actually knows who the candidate is and why they might be interested.

What Passive Candidates Actually Need to Hear

Because passive candidates are skeptical of recruiting messages by default, a message has to deliver something other than "great opportunity, want to talk?" There are four things they need to hear, and each one comes from a different part of the message rather than from a single clever line. Marcus started treating these as requirements rather than aspirations, checking each draft against all four before sending, and that single change did more for his reply rate than any rewording ever did.

Proof of real research. The candidate needs to conclude, in the first two lines, that you did real work to understand them rather than plugging a name into a template. That conclusion comes only from specific references to their actual work, projects, or achievements. Respect for their time. They need to hear that you understand they are happy where they are, that you are not trying to pressure them, but that you found something you thought they would genuinely find interesting. That comes from tone, specifically a tone that treats staying in their current role as a legitimate choice rather than a problem to be solved.

A specific reason they should care. They need to see why this opportunity connects to their work in particular, which comes from showing a genuine gap between what they are doing now and what you are building. Not a better version of their job; a different problem they might find interesting. No pressure. They need permission to walk away, stated plainly: if you are not interested, that is completely fine, and if you want to talk, great. That removes the transactional feel and lets them opt in on their own terms, which is the only way a passive candidate ever enters a process.

Research Before Any Drafting Happens

Marcus's outreach is only ever as good as the research underneath it, and AI cannot do this part for him. AI does not know that a particular engineer just gave a conference talk on migrating a monolith to event-driven services, or that the talk is the genuine hook. Marcus has to find that, and it takes him 15 to 20 minutes per person. He works from public information only: LinkedIn profiles and posts, conference talks, public GitHub activity, company engineering blogs, and published writing. He does not buy contact data, scrape private email, or use anything the candidate did not choose to make public. That boundary is both ethical and practical, since outreach built on information someone never shared reads as surveillance and kills trust instantly.

For each candidate Marcus captures five specific things in a short research note, and they are the same five every time so that the note is comparable across a list. Specific projects and achievements: what did they actually build, and what was the impact? He wants evidence, not adjectives, and the useful form is concrete, along the lines of "led migration to microservices, reducing latency by 40 percent," "built a real-time analytics dashboard serving 500,000 users," or "open-source library downloaded 2 million times." A line like that is a hook. "Experienced backend engineer" is not, because it could describe ten thousand people.

Timeline and career progression: how long have they been in the current role, how long were they in previous ones, is there a pattern, and does that pattern suggest they may be due for a change? Technical depth: which technologies do they know deeply, what are they visibly learning, and what appears to genuinely interest them? Signals of openness: is there anything suggesting they might be receptive right now, such as a LinkedIn update, renewed open-source activity, or conference speaking? Work style and values: from how they describe their work and what they choose to present publicly, what can you infer about how they operate and what they care about, whether that is impact, mentorship, or technical excellence?

Fifteen to twenty minutes per person sounds expensive until you compare it to the cost of forty generic messages that produce four replies. The research note is what Marcus will feed to AI, and the quality of the note caps the quality of the draft, so he refuses to draft from a thin one. When he cannot find a genuine hook after twenty minutes, he does not write a message with a manufactured one. He moves the candidate to a later pass and spends the time on someone he can write to honestly.

Reading Timing Signals Without Being Creepy

When Marcus reaches out matters almost as much as what he says. Certain public signals suggest a passive candidate may be more open than usual: a recent LinkedIn update such as a job change, a new skill, or a relocation; open-source activity resuming after a quiet stretch, which suggests they have time and energy again; a conference talk that hints at wanting more visibility or building a platform for a next move; a job change or reorg at their current company that may leave them due for a bigger change; and unusually long tenure in a single role that may have plateaued. Marcus uses these to prioritize, reaching out first to people showing a signal rather than blasting the whole list at once.

Referencing these signals is not creepy; it is observant. The line between observation and surveillance is whether the signal was offered to the public and whether referencing it feels natural in a professional conversation. "I saw your QCon talk on event-driven migration" is observation, because the talk was a public, intentional act and the candidate published it hoping people would notice. "I noticed you changed your commute" is surveillance, even if the data point exists somewhere. Marcus only references things the candidate clearly chose to publish, and he references them as the reason he is writing, not as proof he has been watching.

The Four-Paragraph Structure for Passive Outreach

Passive outreach has a specific shape, and it is different from what works on active candidates. An active candidate wants to know about the role, so leading with the role is fine. A passive candidate does not care about your role yet, so leading with it guarantees a delete. Marcus's structure is four short paragraphs, each doing one job, and he keeps them in this order because reversing any two of them changes what the message means.

Paragraph one: a genuine compliment plus a specific reference. This is where the research shows. Naming the actual talk, library, migration, or metric demonstrates that real work went into this message and communicates respect for what they have built. Paragraph two: the bridge to your company. This connects their work to yours, and it is where the genuine reason to care lives. The bridge should be a real technical or problem-shaped connection, not a claim that they would love it here.

Paragraph three: respect their status. Say plainly that you understand they may be happy where they are and that you are not assuming otherwise. This paragraph is the one most recruiters cut for length, and it is the one that most reliably separates a message that gets read from one that gets deleted. Paragraph four: the ask, with the pressure removed. Marcus's version reads close to: "If you're open to a quick conversation about this, I'd be grateful. If not, no worries at all, keep doing great work." That makes saying no genuinely easy, which is exactly why people say yes.

Drafting with AI After the Thinking Is Done

Once the research note exists, Marcus uses ChatGPT to turn it into a draft, and his prompt carries all the context the model needs rather than expecting it to guess. He gives it the candidate's specific achievement, the career timeline, the technical focus, and the genuine connection to the role he is filling. Then he states what the message must do: show respect for their current role, make clear he is interested in them as a person rather than as a way to fill a req, and invite conversation without pressure. He closes with a tone brief: genuine, peer to peer, respectful, not salesy, short at four or five paragraphs, no exclamation points and no false urgency. A prompt with that much context gives the model everything it needs to produce something authentic; a prompt without it produces a template.

Marcus never sends the raw AI draft. AI is good at structure and at rephrasing his research into clean prose, but it tends to overstate the match and slip in generic praise that any candidate could receive. So he edits every message: he cuts the line claiming the role was "made for" them, he restores a detail only he would know, and he reads it aloud to check that it sounds like him. Reading aloud is a surprisingly good test, because inauthenticity that survives a silent read rarely survives being spoken. The division of labor is stable: research and judgment from the human, drafting and polish from the AI, final voice from the human again.

Multi-Touch Nurture and the Long Game

Most passive candidates will not reply to a first message, and that is normal rather than failure. Marcus runs a light multi-touch sequence: an initial value-led message, a single follow-up about a week later that adds something new such as a relevant article or a different angle on the role, and then he stops. Two touches and out. He does not send a third or fourth message that simply asks again, because repetition without new value is what makes recruiters feel like spam. Candidates who do not respond go into a nurture list, not a delete pile, and Marcus may reconnect months later when a signal changes or a better-matched role opens.

This respects two real constraints. Repeated unsolicited contact erodes goodwill, and in some channels and jurisdictions it also raises consent and anti-spam considerations, so Marcus keeps email outreach honest about who he is and easy to opt out of, in the spirit of CAN-SPAM, and he treats any candidate data he stores as something to handle carefully under privacy expectations and GDPR where it applies. Value-led patience beats volume, because the goal of passive engagement is a relationship that pays off over a year, not a reply this afternoon.

A Worked Example: Marcus's Rebuilt Sequence

Take a single candidate, an engineer Marcus will call by the placeholder name in his notes, who gave a public talk on cutting database latency and whose GitHub shows recent activity on a caching library. Marcus spends 18 minutes building a research note covering the talk, the library, a four-year tenure at the current company, and a stated interest in mentoring from a blog post. Then he prompts ChatGPT with the achievement, the genuine connection to his team's caching-heavy platform, and his tone brief. The draft comes back structured and clean. He edits it: he removes a sentence overselling the fit, adds a specific reference to the candidate's library, and trims the close to one low-pressure line.

Over a month Marcus runs his old approach and his new approach side by side across his list. The old generic messages reply at roughly 11 percent. The researched, AI-drafted, human-edited sequence with a single value-add follow-up replies at roughly 27 percent, and the replies are warmer, including several "not now, but keep me in mind" responses that seed his nurture list. These numbers are illustrative of the pattern rather than a guarantee, but the direction is consistent: the cost is about 20 minutes of research per candidate, and the return is a reply rate that more than doubles and a pipeline of passive candidates who now recognize his name.

Three Ways Passive Outreach Fails

Generic "you would be great for this" messages. The most common failure is reaching out with generic praise attached to a job description: "I think you'd be a great fit for our open role. We're hiring a senior engineer." It fails because passive candidates know with certainty that you sent that to everyone, and it demonstrates neither research nor genuine interest. It is also self-defeating at scale, since the more of these you send the more you train your target market to ignore your name. The fix is a hard rule: every message includes a specific reference to their actual work or achievements, and if you cannot produce one, you are not ready to send.

Overstating the match. The second failure is making the opportunity sound perfect for them when it might not be, in the register of "this role was literally designed for someone with your background." It fails because passive candidates are skeptical by default and oversell destroys credibility immediately, usually in the exact sentence where it appears. There is a further cost: a candidate who takes the oversell seriously and then discovers the reality has learned something about you that will outlast this req. Honest, realistic outreach reliably gets better response rates than oversold messages, so the correction is to describe the opportunity accurately and let the candidate judge the fit themselves.

Not respecting their status. The third failure is writing as though the candidate must be dissatisfied, in the form of "I know you must be looking for a bigger challenge than what you have now." It fails because you are making an assumption about someone's satisfaction that you have no basis for, and even in the cases where it happens to be true it is presumptuous, which is worse than being wrong. The correction lives in paragraph three of the structure: acknowledge that they may well be happy where they are and treat that as a legitimate position rather than an obstacle.

Practice

Work these in sequence on your own req list. Each exercise produces an input the next one needs.

• Deep research on a passive candidate. Pick a real or hypothetical passive candidate and spend a full 15 to 20 minutes researching their work, projects, career timeline, tech stack, and values. Document what you learn in a written research note rather than holding it in your head.
• Identify your genuine hook. From that note, name one or two things about this person that genuinely connect to what you are building. Write them down as a sentence. If you cannot write the sentence, you do not have a hook yet.
• Draft your outreach. Using the four-paragraph structure and your research note, ask AI to draft the message. Follow the process in order: research, insight, draft, edit. Do not let the model skip straight to a draft with thin context.
• Run the authenticity test. Read the drafted message aloud. Does it sound like you? Does it demonstrate genuine research? Does it respect their status? Edit or revise until it passes all three tests, not two of them.
• Prioritize by timing signal. Look at your target candidates' recent public activity across LinkedIn, GitHub, and conference programs. Identify which ones show a signal of openness, and reach out to those first rather than working the list alphabetically.

Reflection

• Think about your last few successful passive recruits. What actually made them say yes: the research, the timing, or the opportunity itself? What does that tell you about where to spend effort next time?
• How much time do you currently spend researching a passive candidate before writing? What might you learn if you spent twenty minutes instead of five?
• Which timing signal have you not been paying attention to? LinkedIn activity, conference speaking, open-source patterns, internal reorgs? How could you build a habit of noticing it?
• Where does your outreach most often turn generic: the opening, the bridge to your company, or the close? What would a specific version of that section look like?
• How many touches do you currently send before giving up, and are the later ones adding value or just repeating the ask?

Glossary

• Passive candidate. Someone who is not actively looking for a job. They are employed and satisfied enough not to be searching, which requires a different outreach approach than active candidates.
• Research depth. The amount of work you put into understanding a candidate's background, work, interests, and values. Passive outreach requires deeper research than active-candidate outreach.
• Timing signal. A public indicator that a passive candidate might be open to conversation, such as a LinkedIn update, renewed open-source contribution, or conference speaking.
• Authenticity. The quality of sounding like a real person who did real research and genuinely cares. Passive candidates are unusually sensitive to inauthenticity.
• Pressure-free messaging. Outreach that makes it easy for a candidate to say no without guilt, giving them a clean way to opt out of the conversation.
• Research note. The short written record of what you found about a candidate, structured the same way every time, which becomes the context you hand to AI before drafting.

Related Lessons

Passive engagement sits at the intersection of sourcing, drafting, and privacy, and several lessons extend it in each direction.

• AI-Assisted Outreach: Templates, Personalization, and Quality covers the general drafting mechanics that passive outreach specializes, including where templates legitimately help and where they betray you.
• Scenario Practice: Draft, Review, and Improve runs a full passive-outreach cycle end to end alongside a rejection and a counter-offer, with the verification checklist applied to each.
• Avoiding Generic or Manipulative Messaging goes deeper on the first two anti-patterns here, especially the difference between persuasive and manipulative framing.
• Research Synthesis: Building Candidate Context from Multiple Sources is the discipline behind the research note, including how to reconcile sources that disagree.
• Privacy Boundaries: Data Sharing, Tool Selection, and Compliance covers what candidate information you may put into an AI tool and what you may not, which matters as soon as your research note leaves your notebook.
• Personalization at Scale: When AI Enables Better Communication addresses the tension between depth and volume that passive sourcing creates.

Closing

Passive candidate recruitment is where the best candidates often hide. They are not actively looking precisely because they are successful and valuable, and that same fact makes them worth pursuing. Their selectivity is good news rather than an obstacle: it means the competition is not really other companies, it is the dozen generic messages already sitting unread in their inbox. Most of your competitors will not do deep research or write authentic outreach, because it is slower. If you do the work, you stand out by default.

Make it concrete this week. Identify your top five passive candidates and spend 15 to 20 minutes researching each one. Write down the genuine hook for each. Then draft and send one personalized message per day. By the end of the week you will have sent five authentic messages to people who are accustomed to receiving spam, and that authenticity is what produces responses. Passive candidates represent a long-term recruiting advantage rather than a way to fill this month's req, and the pipeline you build now is the one that pays out next year.

Key Takeaways

• Passive outreach earns a reply, it does not sell a role. The candidate has decided nothing, so the first message must prove genuine attention rather than pitch a job description. A reply, even a "not now," is the win.
• Four things must be present in every message. Proof of real research, respect for their time, a specific reason this connects to their work, and no pressure. Check all four before sending rather than hoping the message conveys them.
• Research caps quality, and it is the human's job. Fifteen to twenty minutes of public-only research per candidate produces the specific hook that makes a message land. AI cannot know what makes someone interesting, so the research note must come first.
• Structure the message in four moves. Genuine compliment with a specific reference, bridge to your company, explicit respect for their current situation, and a low-pressure ask that makes saying no easy.
• Reference only what the candidate published, and frame it as a reason you are writing. A public talk or repo is observation. Anything the person did not choose to share is surveillance and destroys trust instantly.
• Prioritize by timing signal. LinkedIn updates, resumed open-source activity, conference speaking, internal job changes, and long plateaued tenure all suggest openness. Work those candidates first instead of messaging the whole list at once.
• Use AI for structure and drafting, keep judgment and voice human. Give the model your research note and a clear tone brief, then edit every draft to cut overstatement, restore real detail, and make sure it sounds like you.
• Be honest rather than enthusiastic. Do not overstate how perfect the match is or assume the candidate is unhappy. Honest, realistic outreach gets better response rates than oversold messages, and it survives contact with reality.
• Two touches with new value, then nurture, not nag. A first message and one follow-up that adds something fresh respects the candidate and stays on the right side of anti-spam and privacy norms like CAN-SPAM and GDPR. Repetition without value is spam.
• Measure the trade, not the message. About 20 minutes of research per candidate that roughly doubles reply rate and builds a warm nurture list is a strong return, because passive engagement pays off over a year, not an afternoon.

Frequently Asked Questions

Twenty minutes per candidate does not scale. How do I run this against a list of eighty? You do not, and that is the point. The arithmetic that matters is replies per hour, not messages per hour. Forty generic messages producing four replies is a worse use of a day than twelve researched messages producing three, because the researched replies are warmer and several of them become nurture-list relationships rather than one-off declines. Practically, prioritize by timing signal so your deep research goes to the people most likely to be receptive right now, and let the rest of the list wait for a signal rather than receiving a shallow message today.

What if I research someone for twenty minutes and cannot find a genuine hook? Then do not send. A manufactured hook is worse than no message, because passive candidates recognize it instantly and it costs you the ability to write to that person credibly later. Move them into a nurture list and revisit when something changes: a talk, a published post, a new project, a reorg at their company. The absence of a hook is often information, telling you either that this person's work is not actually adjacent to your problem or that their public footprint is too thin to write from, and both are reasons to spend the time elsewhere.

Where exactly is the line between observant and creepy? Two tests, and a reference has to pass both. First, did the candidate publish this deliberately for a professional audience? A conference talk, a public repository, a blog post, and a LinkedIn update all qualify. Second, would it feel natural to mention in a first professional conversation? If a reference would make someone wonder how you found that out, it fails regardless of whether the data was technically available. Marcus also frames every reference as the reason he is writing rather than as evidence of how much he knows, which keeps the message reading as interest rather than as a dossier.

Should I disclose that I used AI to draft the message? The judgment that matters is yours, and the message should be true. The research is human, the hook is human, and the final edit is human, so what the AI contributed is phrasing rather than substance. What genuinely matters is that every factual claim in the message is one you verified and every sentence is one you would defend if quoted back to you. The failure mode to avoid is not undisclosed drafting assistance, it is sending a message containing an achievement the model invented or a claim about your company you have not checked.

How long should I wait before the follow-up, and what should it contain? About a week, and it has to add something new. A follow-up that repeats the ask in different words is the message that converts a neutral non-response into an active annoyance. New value can be small: a relevant article, a different angle on the problem your team is solving, or an update that makes the opportunity more concrete. After that second touch, stop. Move the candidate to the nurture list and reconnect months later when a signal changes or a better-matched role opens, which is a genuinely different conversation rather than a third ask.

My AI drafts keep sounding oversold no matter how I prompt. What is going on? Two causes, usually together. The first is a thin research note: with no specific material to work from, the model fills the space with enthusiasm, because enthusiasm is what generic recruiting text is made of. The second is a missing tone constraint. State it explicitly and negatively as well as positively: peer to peer, not salesy, no exclamation points, no false urgency, no claims that the role was made for them. Then edit anyway. The overselling sentence is the one the model reaches for when it runs out of substance, and cutting it is often the single edit that makes the message sound human.