ConnectDevs

Category: AI Candidate Sourcing

Tactics and tools for finding and reaching qualified candidates, including passive talent.

  • Recruiter InMail Response Rate: Benchmarks and 9 Ways to Raise It

    Recruiter InMail Response Rate: Benchmarks and 9 Ways to Raise It

    Recruiter InMail Response Rate: Benchmarks and 9 Ways to Raise It

    Your recruiter InMail response rate decides how many candidates you actually talk to, and for most teams it is lower than they think. LinkedIn’s own guidance for Recruiter accounts is a 13% response rate, and plenty of recruiters sit below it. This guide shows what a good rate looks like, where the numbers come from, and nine changes that raise replies without sending more messages. It also covers cold outreach to candidates beyond LinkedIn, because the best teams no longer rely on one channel. Every statistic below links to its source, and vendor-reported data is labeled.

    TL;DR

    • A good InMail response rate is roughly 15-25%, and LinkedIn expects at least 13% on Recruiter accounts.
    • Cold email to candidates replies far lower, around 5-6% per message in one 2026 vendor dataset.
    • The biggest levers are targeting, a specific opener, short copy, two to three follow-ups, and a second channel.
    • Track positive replies, not just replies, or you will optimize for the wrong thing.

    Table of Contents

    What is a good InMail response rate?

    A good recruiter InMail response rate is about 15-25%, in line with the 10-25% range for personalized InMail that daily.dev’s roundup reports (a secondary, vendor-published figure). Below 13% is a warning sign, because LinkedIn’s own rule says Recruiter and Recruiter Lite users should hold a 13% response rate across 100 or more InMails in every 14-day period (as quoted by Pin from LinkedIn Talent Solutions). Above 30% usually means a tight, specific talent pool and short, personal messages.

    Context matters. A senior staff engineer gets many more messages than a regional operations manager, so the same copy earns different results. Compare yourself to your own role types first and to the market second.

    Recruiter outreach benchmarks by channel

    Public InMail benchmarks are thin and mostly come from vendors, so treat them as ranges. The cleanest recent dataset is Pin’s 2026 report, which analyzed 4 million+ messages across 1,500+ organizations between June 2025 and May 2026. It is vendor-reported data, but the sample is large and the method is stated (Pin recruiting outreach benchmarks 2026).

    Channel Reply rate (Pin, 2026) What it tells you
    Automated email 4.96% Generic sequences get the lowest replies.
    Recruiter-written email 6.31% Human edits help, but cold email is still a low-reply channel.
    LinkedIn messages 17.08% The in-platform inbox is where candidates expect recruiters.
    Email + LinkedIn, 2-step sequence 45.76% Versus 19.73% for email-only in the same report.

    Two other data points are worth knowing. Pin’s matched-candidate analysis of 165,000+ people contacted on both channels found that 78% of repliers answered only on LinkedIn, 17% only on email, and 5% on both (Pin channel comparison). Gem, which analyzed 6.2 million email sequences for its 2026 report, publishes its full benchmarks behind a download (Gem 2026 email outreach benchmarks).

    Why most InMails get ignored

    Candidates, especially engineers, see a steady stream of near-identical messages. Most fail for the same four reasons.

    • Wrong person. The profile never matched the role closely enough.
    • No reason. The message could be sent to anyone, so it reads that way.
    • Too long. A wall of company history buries the ask.
    • One shot. Many candidates simply miss the first message, and the recruiter never follows up.

    Notice that none of these is a subject line trick. Fix targeting and relevance first, and the formatting tweaks matter later.

    9 ways to improve your InMail response rate

    1. Send fewer, better-matched messages

    Tighter lists beat bigger lists. daily.dev’s roundup of InMail data reports that campaigns of about 50 recipients get roughly three times the replies of mass sends, though that is a secondary compilation, so test it on your own roles (daily.dev InMail statistics). Before you write a word, ask whether you would pass this profile to the hiring manager. If the answer is “maybe,” skip it.

    2. Open with a reason specific to them

    Name something real: a project, a repo, a talk, a stack choice, a promotion. One sentence of genuine relevance does more than three sentences about your company. Even basic personalization shows up in the data. In Pin’s email dataset, messages using the candidate’s first name replied at 5.13% versus 2.61% without it.

    First names are the floor, not the goal. Specifics are what separate you from the other twenty recruiters in that inbox.

    3. Keep it short

    daily.dev’s compilation, citing LinkedIn research, reports that InMails under 400 characters are 22% more likely to get a response. For email, Pin found the best reply rate (5.46%) at 150-199 words, while messages under 50 words did worst at 2.43%. So short is good, but not empty. Give enough to make a decision.

    4. Put the deal terms up front

    Engineers and other in-demand candidates decide fast. State the stack, work model, level and, where you can, the pay range in the first lines. Hiding these details forces a reply just to learn basics, and many candidates will not bother.

    5. Write a plain, short subject line

    Pin’s data shows subject lines of five to six words performed best at a 5.35% reply rate. Skip clickbait and “Quick question.” Name the role and the reason, such as “Staff backend role, your Kafka work.”

    6. Follow up two or three times

    Most replies do not come from the first message. In Pin’s data, the first three touches produced 93.2% of all replies, and a fourth added little (97.7% cumulative). Message 2 alone drew 33.7% of replies.

    So plan three touches, add new information each time, and stop. Candidates who reply to touch four are rare, and you risk annoying people you may want later.

    7. Add a second channel

    This is the biggest lift in the dataset. Pin reports 45.76% reply rate for two-step email plus LinkedIn sequences versus 19.73% for email-only, a 2.3x lift. Treat InMail as one touch in a sequence, not the whole plan.

    If you already hit your InMail credit limits, email gives you a way to reach candidates outside LinkedIn. Our guide to finding passive candidates beyond LinkedIn covers where to find them.

    8. Send on mid-week days

    In Pin’s email data, Thursday (5.16%) and Wednesday (5.14%) led, and Saturday (3.90%) trailed. The gaps are small, so do not rebuild your workflow around them. Just avoid weekends and test your own audience.

    9. Reply fast and make the next step easy

    Candidates reply quickly or not at all. Pin reports 74.8% of replies arrive within 24 hours, with a median of about four hours. If a reply sits in your inbox for two days, you are wasting the work you just did.

    Offer one concrete next step, like a 15-minute call or an async screen link, rather than “let me know your thoughts.”

    Before and after: a weak vs strong message

    The table below is an illustrative example, not measured data.

    Element Weak version Stronger version
    Subject Exciting opportunity! Senior data engineer, your Spark work
    Opener I came across your profile and was impressed. Your talk on streaming pipelines at a recent meetup matched what our team is building.
    Details Fast-growing company, great culture. Senior level, Python and Spark, remote in EU time zones, stated pay range.
    Ask Let me know if you are interested. Open to a 15-minute call Wednesday or Thursday?
    Follow-up None Two short follow-ups, each with new information

    How to measure it properly

    Total reply rate hides a lot. “Not interested” and “please remove me” count as replies, yet they do not fill a pipeline. Track three numbers per role: reply rate, positive reply rate, and replies that became a screen.

    Also segment by role family and seniority. A 12% rate on principal engineers might beat a 25% rate on junior roles. Review weekly, change one variable at a time, and give each test at least 100 sends so the result means something.

    Manual follow-ups and reply tagging eat recruiter hours, which is why many teams automate that layer. Our post on recruiter productivity shows where the time goes, and LinkedIn Recruiter alternatives compares tools if InMail credits are your bottleneck.

    FAQ

    What is a good recruiter InMail response rate?

    Aim for 15-25%. LinkedIn expects at least 13% for Recruiter and Recruiter Lite users over 100+ InMails per 14 days. Rates above 30% usually come from narrow, well-matched talent pools.

    How do I improve my InMail response rate?

    Start with targeting, then write a short message with a specific reason for contacting that person and the key role details up front. Follow up two to three times and add a second channel such as email. Track positive replies so you know what is working.

    How long should a recruiter InMail be?

    Short. A compilation citing LinkedIn research reports that InMails under 400 characters are 22% more likely to get a reply. Aim for three or four short sentences with the role, the reason, and one clear ask.

    How many times should I follow up with a candidate?

    Two or three follow-ups is usually enough. In Pin’s 2026 vendor data, the first three touches generated 93.2% of replies. After that, returns drop quickly.

    Is email or InMail better for cold outreach to candidates?

    In Pin’s 2026 data, LinkedIn messages replied at 17.08% against 4.96-6.31% for email. The channels reach different people, though, and combining them lifted two-step sequence replies to 45.76% from 19.73% for email alone.

    What is the best day to send recruiter outreach messages?

    Wednesday and Thursday led in Pin’s email data, and Saturday was weakest. The differences are modest, so message quality matters far more than send day.

    Why is my InMail response rate low?

    Usually the cause is weak targeting or a generic message, not the subject line. Check whether the profiles truly match the role, then check whether your opener gives a specific reason to reply. Fixing those two beats any formatting tweak.

    Conclusion

    Raising your recruiter InMail response rate is mostly a discipline problem: match tighter, say something specific, keep it short, follow up, and use more than one channel. Start with two changes this week, measure positive replies, and add the rest as you go.

    If the follow-ups and reply sorting are what slow you down, ConnectDevs Pilot sends multi-touch campaigns from your own Gmail, stops follow-ups when a candidate replies, and sorts replies into Interested, Meeting, Follow up, or Not now. To find better-matched candidates before you write, see AI sourcing.

  • X-Ray Search for Recruiters: Google Strings That Still Work in 2026

    X-Ray Search for Recruiters: Google Strings That Still Work in 2026

    X-Ray Search for Recruiters: Google Strings That Still Work in 2026

    X-ray search for recruiters means using Google’s site: operator to search inside LinkedIn, GitHub, Stack Overflow and other sites from outside their own search boxes. For years it was the cheapest way to find passive candidates. Since early 2024 it has been less reliable on LinkedIn, but it still works well on developer sites.

    This guide gives you copy-paste strings, explains what changed, lists the mistakes that return empty results and shows when it is time to stop typing operators and use AI sourcing instead.

    TL;DR

    • X-ray search is a Google query pattern: site: plus quoted keywords, a location and exclusions.
    • It works best on GitHub and Stack Overflow. On LinkedIn it returns thinner results than it did before 2024.
    • Most failures come from small syntax errors: a space after site:, a lowercase OR or missing parentheses.
    • X-ray finds names, not contact details or outreach. Plan for a separate enrichment and outreach step.
    • Switch to AI sourcing when you run the same search across several platforms every week.

    Table of Contents

    What is X-ray search?

    X-ray search is a way to search a specific website through Google by adding the site: operator to a query. Instead of using LinkedIn’s search, you ask Google to return only pages from linkedin.com/in that match your keywords. The same idea works for GitHub profiles, Stack Overflow user pages, university sites and public PDF resumes.

    It is a form of Boolean search. If you need a refresher on operators like AND, OR and NOT, start with our guide to Boolean search for recruiters. X-ray adds one idea on top: the site restriction.

    The basic formula is:

    site:[domain or path] "keyword 1" "keyword 2" "location" -exclusion

    Google operators recruiters use

    Google’s own search help page documents fewer operators than recruiters use in practice, so test each string before you rely on it. These are the ones that matter:

    Operator What it does Example
    site: Limits results to a domain or path site:github.com
    " " Matches an exact phrase "machine learning"
    OR Matches either term (use capitals) ("Django" OR "FastAPI")
    - Excludes a term -jobs -directory
    intitle: Term must appear in the page title -intitle:repositories
    inurl: Term must appear in the URL -inurl:status
    filetype: Finds a file type such as PDF filetype:pdf "resume"
    * Wildcard for a missing word "senior * engineer"

    Some old operators no longer work. Pin’s X-ray search guide reports that Google retired related: in 2023 and cache: in 2024. If a string copied from an old blog post returns nothing, an outdated operator is a likely cause.

    Copy-paste X-ray strings

    Swap the bracketed parts for your own role, skills and location. These patterns are adapted from the examples in Pin’s guide and 100Hires’ LinkedIn X-ray guide.

    GitHub: developers by skill and city

    site:github.com "Python" ("Django" OR "FastAPI") "Berlin" -intitle:repositories

    The -intitle:repositories exclusion removes repository listing pages, so you see more user profiles. Our full walkthrough on finding passive candidates beyond LinkedIn covers what to do once you land on a profile.

    Stack Overflow: people who answer questions in your stack

    site:stackoverflow.com/users "java" "toronto"

    User pages show the topics someone answers about, which is useful evidence of depth in a specific technology.

    LinkedIn: public profiles (reduced results)

    site:linkedin.com/in ("software engineer" OR "backend engineer") "austin" -jobs -directory

    Expect fewer and less complete results than the same search in LinkedIn Recruiter. More on why below.

    Public resumes: PDF files

    filetype:pdf "software engineer" "Python" "Chicago" -jobs -template -sample

    The exclusions remove job ads and resume templates. Resumes found this way can be years old, so check the candidate’s current role before you reach out.

    University project pages

    site:*.edu filetype:pdf "resume" "computer science" "machine learning"

    Useful for early-career roles and research-heavy teams.

    How to build your own string in five steps

    1. Pick the platform. Choose where your candidates leave public evidence. Engineers leave code on GitHub and answers on Stack Overflow. Designers and marketers are more likely to appear on portfolio sites.
    2. List the title variants. Candidates write “Software Engineer”, “SDE” or “Backend Developer”. Put the variants in one parenthesis group joined with OR.
    3. Add two or three skills. Use the terms a candidate would write on their own profile, not the wording from your job post.
    4. Add a location. A city name filters well. For remote roles, leave it out and filter later.
    5. Exclude noise. Add -jobs, -intern or -recruiter one at a time and check the results after each change.

    Why LinkedIn X-ray got worse

    Sourcing experts documented a change in early 2024: Google began indexing fewer fields from public LinkedIn profiles. According to Pin’s guide, Google can no longer see work history, education, skills or most location data on many profiles. A search that once matched a skill in the middle of a profile now only matches the headline and a few visible fields.

    Public profiles can also take weeks or months to refresh in search engines, so a title you see in Google may be out of date. LinkedIn also does not publish a feed of changes for Google to follow. Treat every X-ray result as a lead to check, not as a confirmed current profile.

    If you also pay for LinkedIn Recruiter, see our comparison of LinkedIn Recruiter alternatives for cheaper options.

    Common mistakes that return empty results

    • A space after the colon. site: github.com breaks the operator. Write site:github.com.
    • Lowercase OR. Use OR in capitals.
    • Missing parentheses. Without them, OR can scramble the logic of the whole query.
    • Job-post language. “Seeking an experienced engineer” is not how people describe themselves. Search for the terms they actually write.
    • Too many exclusions. Each minus term can hide real candidates. Add them gradually.
    • Typos in the path. linkedin.com/il returns nothing. The profile path is /in.

    Long strings can also be cut short. 100Hires reports that Bing limits queries to 10 terms, so the end of a long string may be ignored. Keep strings short and test in Google first.

    Limits and compliance

    X-ray search is useful, but it has real limits:

    • Exact-match dependence. It misses people who describe the same skill in different words.
    • One platform at a time. You run, review and de-duplicate each site’s results by hand.
    • Partial coverage. Pin’s own index analysis, a vendor figure, suggests fewer than 13% of software engineering profiles have a public GitHub. Many strong engineers never appear in a GitHub X-ray.
    • No contact data and no workflow. You get a name and a link. Email lookup, outreach and tracking are manual. Read our guide on recruiter outreach response rates before you start messaging.

    Public does not mean free to use however you like. Check the terms of each platform you search, and follow privacy rules such as GDPR if you store candidate data. Ask your legal or compliance contact before you build a database from scraped results.

    X-ray search vs Boolean search vs AI sourcing

    X-ray search Boolean in a sourcing tool AI sourcing
    Where you search Google, one site at a time Inside one platform Across a profile database
    How you query Operators and exact phrases Operators and filters Plain-language criteria
    Handles synonyms Only if you list them Only if you list them Usually, by design
    Contact data No Sometimes Often through enrichment
    Cost Free Tool subscription Tool subscription
    Best for Niche, one-off searches Precise control Repeat searches at volume

    For a deeper look at the trade-off, read our comparison of Boolean search vs natural language search.

    When to switch from X-ray to AI sourcing

    Keep X-ray in your toolkit for rare skills and quick checks. Move to an AI sourcing tool when:

    • you repeat similar searches every week for different roles;
    • you spend more time cleaning results than contacting people;
    • you need email addresses and a place to track replies; or
    • your candidates are not on GitHub or Stack Overflow.

    ConnectDevs Scout searches a published 800M+ public-profile graph and returns ranked evidence for recruiter review. Missing emails can be filled in with candidate enrichment. Plans start free, with paid tiers listed on the pricing page. You still review every shortlist yourself.

    FAQ

    What is X-ray search in recruiting?

    X-ray search is a Google search that uses the site: operator to find profiles on one website, such as GitHub or LinkedIn, using keywords, locations and exclusions.

    Does LinkedIn X-ray search still work in 2026?

    It still returns some public profiles, but results are thinner than before 2024 because Google indexes fewer profile fields. Use it for leads and verify them.

    Is X-ray search free?

    Yes. Google search costs nothing. Your costs are time and the tools you need for email lookup and outreach.

    Is X-ray search against LinkedIn’s rules?

    Searching Google for public pages is different from scraping LinkedIn. Automating collection or copying profile data at scale can breach platform terms and privacy law. Check the current terms and ask your legal team.

    Why did my X-ray search stop working?

    The usual causes are a retired operator, a syntax error such as a space after site:, too many exclusions, or Google’s reduced LinkedIn indexing.

    Can AI write X-ray search strings?

    Yes, a chat assistant can draft strings quickly. Test each one, because assistants sometimes use retired operators or break the parentheses.

    Conclusion

    X-ray search is still worth learning. It costs nothing, it works well on GitHub and Stack Overflow, and it teaches you how candidates describe themselves. Use it for narrow, one-off searches, and accept that LinkedIn results are now partial. When the same work repeats weekly, a tool that searches intent instead of exact strings will save you the hours.

    Tired of rebuilding strings for every role? Explore AI candidate sourcing and run your next search on a live role.

  • ChatGPT for Recruiters: 12 Prompts That Work and 5 Things It Can’t Do

    ChatGPT for Recruiters: 12 Prompts That Work and 5 Things It Can’t Do

    ChatGPT for Recruiters: 12 Prompts That Work and 5 Things It Can’t Do

    ChatGPT for recruiters works best as a drafting and thinking partner, not as a sourcing or screening system. It can write a job description in seconds, suggest alternative job titles and tighten an outreach email. It cannot search for candidates, keep an audit trail or tell you who changed jobs last week.

    This guide gives you twelve prompts you can paste today, shows where each one saves time, and lists the limits that matter before you rely on it. It also covers candidate data privacy, which is where most teams make their first mistake.

    TL;DR

    • Use ChatGPT for drafting, research and structure: job descriptions, search strings, outreach, interview questions and scorecards.
    • Do not use it to make or justify screening decisions, and do not paste candidate names or contact details into a consumer plan.
    • It has no candidate database, no sending, no ATS link and no bias audit trail. Pair it with tools that do.
    • Edit everything it writes. Generic AI outreach is easy for candidates to spot.

    Table of Contents

    Where ChatGPT helps recruiters most

    ChatGPT is useful in recruiting wherever the task is writing, summarizing or structuring, and where a human reviews the result. SHRM’s guidance on generative AI for sourcing names four main uses: candidate research, refining job descriptions, composing outreach and crafting interview questions. AIHR’s guide adds preparation, screening support and onboarding to the list, and covers the same boundaries in its 20-prompt ChatGPT recruiting guide.

    Adoption is already common. SHRM’s 2025 research, as cited by Pin, found 69% of HR professionals use AI to support recruiting, and 89% of those say it saves time or improves efficiency. The same source cites a SHRM finding that 19% of organizations using AI in hiring say their tools overlooked or screened out qualified applicants. Both numbers matter: time savings are real, and so is the risk of quiet errors.

    12 prompts that work

    Replace the bracketed text with your own details. Give the model your real context, because vague inputs produce generic output.

    Intake and job descriptions

    1. Intake questions. “I’m a recruiter opening a [role] at a [company stage] company. Write 10 intake questions for the hiring manager that surface must-haves, nice-to-haves, deal-breakers and the first-90-day outcomes.”
    2. Job description draft. “Draft a job description for [role]. Lead with the problem the hire will solve. Include stack, pay range [range], process steps and what we do not require. Keep it under 450 words and avoid cliches.” For structure, see our technical job description template.
    3. Bias check. “Review this job description for exclusionary phrasing and unnecessary requirements. List each issue and suggest a neutral rewrite.”

    Sourcing research

    1. Alternative titles. “List 15 job titles, including less obvious ones, used for a person who does [work]. Group them by seniority.”
    2. Skill adjacency. “Which skills or tools commonly appear alongside [skill] in resumes for [role]? Separate core from adjacent.”
    3. Boolean draft. “Write a Boolean search string for [role] with these must-have skills [list] and these exclusions [list]. Explain each operator.” Always test and trim the result. Our Boolean search guide shows what good strings look like.
    4. Company target list. “Suggest 20 company types that employ people with experience in [domain]. Do not name individuals.” Verify every company before using the list.

    Outreach and screening support

    1. Outreach draft. “Write a 90-word email to a [role] about [role summary]. Reference this verified detail: [one fact you found]. End with one low-pressure question.”
    2. Follow-up variations. “Give me three follow-up messages for a candidate who has not replied after five days. Each should add new information, not just nudge.”
    3. Interview questions. “Create 8 structured behavioral and technical questions for [role] that map to these competencies [list]. For each, give a strong, average and weak answer description.”
    4. Scorecard. “Turn these competencies into a 1-4 scorecard with behavioral anchors for each score.” Our guide to AI candidate screening software explains why structured criteria beat gut feel.
    5. Candidate update email. “Write a short, respectful rejection email for a candidate who completed [stage]. Offer one specific piece of feedback placeholder for me to fill in.”

    Two habits improve every prompt. First, state the audience and tone. Second, ask for the output in a fixed format, such as a table or numbered list, so you can compare versions quickly.

    5 things ChatGPT can’t do

    A general chatbot cannot replace a recruiting system because it lacks data, workflow and records. Five gaps show up quickly in daily work.

    1. It has no candidate database. It cannot search for available people, show who recently changed roles or confirm that a profile exists. Anything it states about a named person should be treated as unverified.
    2. It does not send or track outreach. It drafts messages, but it does not manage sequences, replies or deliverability. Pin reports that 61.9% of replies in its sample of 59,000+ outreach sequences arrived more than 24 hours after the first message, which shows why follow-up timing matters (vendor data, not independent research).
    3. It does not connect to your ATS or CRM. Work happens in a browser tab, so notes and decisions stay outside your pipeline.
    4. It has no audit trail. If you ask it to summarize or rank candidates, there is no record of the criteria it weighted. That is a problem if a candidate or regulator asks how a decision was made.
    5. It can sound confident and be wrong. AIHR warns that it may misuse technical or HR terms, and Pin notes it can produce plausible but false details. Review every fact.

    The practical rule: use ChatGPT to make a recruiter faster at thinking and writing, then use purpose-built systems to find, contact and track people.

    Candidate data and privacy

    Do not paste candidate names, emails, phone numbers or full resumes into a consumer chatbot plan. Terms differ by plan and change over time, so read the current data-use and retention terms for the plan you use. Pin’s summary says free and individual plans lack a data processing agreement while business plans include one, but confirm this directly with the provider and with your legal team.

    A safe default is to remove identifying details first. Replace names with “Candidate A,” strip contact information and paste only the parts you need, such as a list of skills or a project description. If your company has an approved AI tool, use that instead of a personal account.

    Compliance and bias

    If AI influences who advances, a growing number of rules may apply. New York City, for example, requires notice and independent bias audits for automated employment decision tools. Drafting an email is different from ranking candidates, but the line blurs fast when a recruiter asks a chatbot who looks strongest.

    Keep these guardrails in place:

    • Do not use ChatGPT to decide who to reject. SHRM quotes one practitioner saying these tools should “screen people in, not screen people out.”
    • Keep a human reviewer on every decision and record the criteria you used.
    • Tell candidates when AI is part of your process where local rules require it.
    • Redesign take-home tests that a chatbot can complete on its own.

    When to move beyond a chatbot

    Drafting is rarely the bottleneck for long. Teams usually hit a ceiling when they need to find people they do not already know, enrich records with current contact routes and keep a trail of what happened. That is where a sourcing system helps. ConnectDevs Scout turns a role brief into a structured search across the 800M+ public-profile graph the company describes, then presents ranked evidence for a recruiter to review. It is built for finding candidates, which a chatbot cannot do.

    If you are weighing tools, our guide to what AI recruiting agents actually automate shows where automation is reliable today. For tactics that do not depend on a chatbot, see how to find passive candidates beyond LinkedIn. Published ConnectDevs pricing starts with a free plan, so you can test it on a live role.

    FAQ

    Can recruiters use ChatGPT?

    Yes, for drafting, research and structuring work. Avoid entering candidate personal data into consumer plans, and do not use it to make hiring decisions.

    What are the best ChatGPT prompts for recruiters?

    The most useful prompts cover intake questions, job descriptions, alternative job titles, outreach drafts, structured interview questions and scorecards. Include role context, tone and output format in every prompt.

    Can ChatGPT source candidates?

    No. It has no candidate database and cannot search profiles. It can suggest titles, skills and Boolean strings that you run in a sourcing tool.

    Is it safe to paste resumes into ChatGPT?

    Only after removing names and contact details, and only on a plan whose data terms your company has approved. Many companies prefer an enterprise-approved AI tool for this reason.

    Will ChatGPT replace recruiters?

    It automates parts of the writing and research work, but judgment, persuasion and accountability stay with people. We cover the split in AI recruiter vs human recruiter.

    Is using ChatGPT in hiring legal?

    Using it to write a job description is low risk. Using AI to score or reject candidates can trigger audit, notice and record-keeping rules in some places. Check local requirements with counsel.

    Conclusion

    Treat ChatGPT as a fast junior writer who needs supervision. Use the twelve prompts to remove drafting work, keep candidate data out of unapproved tools and send decisions through a process you can explain. Then add systems that do what a chatbot cannot: find people, enrich records and keep a trail.

    Ready to go from drafting to finding? Explore AI candidate sourcing and test a live role on the free plan.

  • How to Source Developers on GitHub: A Recruiter’s Step-by-Step Guide

    How to Source Developers on GitHub: A Recruiter’s Step-by-Step Guide

    How to Source Developers on GitHub: A Recruiter’s Step-by-Step Guide

    Knowing how to source developers on GitHub gives you a pool of engineers who are not answering recruiter messages on LinkedIn. It also gives you proof of work, which a resume never will. The catch is that GitHub is built for code, not hiring. You need the right search qualifiers, a way to read profiles without writing code, and clear rules on what you can and cannot do with the data.

    This guide covers all three. It goes deeper than our overview of how to find passive candidates beyond LinkedIn, which only touches GitHub briefly.

    TL;DR

    • Search GitHub for repositories first, then find the people behind them. It beats searching profiles alone.
    • Stack qualifiers such as language:, location:, followers: and repos: to cut thousands of profiles to a shortlist.
    • Judge a developer on recent, relevant, collaborative work. Raw star counts and green squares mislead.
    • GitHub’s Acceptable Use Policies bar using its data for spam, and GDPR adds duties when you contact people. Personalize every message and keep it low-volume.
    • AI sourcing helps with discovery and ranking. A human still has to read the work and write the message.

    Table of Contents

    Why GitHub works for developer sourcing

    GitHub sourcing means finding developers through their public code, contributions and collaboration instead of their job-board profile. You see what they build, how they talk to other engineers, and what they care about.

    The pool is huge. GitHub’s Octoverse 2025 report says the platform passed 180 million developers and 630 million repositories. It also reports TypeScript overtook Python as the most used language in August 2025.

    Size creates its own problem. Most accounts are not candidates, and most profiles are thin. Your job is to filter hard and read closely.

    Who you will and will not find

    GitHub skews toward open-source contributors, front-end and back-end engineers, infrastructure and data people. Engineers who work mostly in private company repos look quiet here. Many strong developers have almost nothing public.

    So treat GitHub as one channel. A thin profile is not a negative signal.

    Step 1: Search with GitHub qualifiers

    Search qualifiers are keywords such as language: or location: that filter GitHub results to match specific criteria. Use them in the GitHub search bar, then switch the result type to Users or Repositories.

    These user qualifiers come from GitHub’s documentation on searching users:

    Qualifier What it does Example
    type:user Personal accounts only, no organizations type:user
    language: Matches the languages used in the person’s repositories language:javascript location:russia
    location: Matches the location in the profile repos:1 location:iceland
    followers: Filters by follower count followers:>=1000
    repos: Filters by number of repositories repos:10..30
    created: Filters by account creation date created:<2011-01-01
    in:name / fullname: Finds a person by name fullname:nat friedman
    in:email Searches public email addresses data in:email

    Combine them. A starting query for a Go engineer in Berlin could look like this:

    language:go location:berlin followers:>20 repos:>5 type:user

    Limits you should know

    • Location is free text. People write “Berlin”, “Berlin, Germany” or “DE”. Run several variants.
    • Email search is restricted. GitHub states that for privacy reasons you cannot search by email domain name.
    • Results are capped. The search API documentation says it returns up to 1,000 results per search. Narrow your query instead of paging forever.
    • Followers are a weak proxy. A senior infrastructure engineer may have 12 followers and a brilliant career.

    Step 2: Start from repositories, not profiles

    Repository-first sourcing means searching for projects that match your tech stack, then reviewing the people who maintain and contribute to them. It finds stronger candidates than profile search, because the work is the filter.

    Repository qualifiers, per GitHub’s repository search docs:

    Qualifier Use it to Example
    topic: Find projects tagged with a technology or domain topic:kubernetes
    language: Limit to one language language:rust
    stars: Filter by popularity stars:10..500
    pushed: Keep only recently active projects pushed:>2026-01-01
    archived: Drop dead projects archived:false
    forks: Spot projects others build on forks:>=5
    good-first-issues: Find projects welcoming new contributors good-first-issues:>2 javascript

    A practical query: topic:kubernetes language:go stars:10..500 pushed:>2026-01-01 archived:false. Update the date to within the last six months when you run it.

    From a repository to candidates

    1. Open a promising repo and check the Contributors list on the right.
    2. Skim the Pull requests tab for merged work by people outside the core team.
    3. Look at who reviews code. Reviewers are often your senior hires.
    4. Open each candidate profile and run the checks in the next step.

    Mid-popularity projects work best. Famous projects attract recruiters from everywhere. A repo with 200 stars and three active maintainers is far less crowded.

    Step 3: Read a profile without coding skill

    Reading a GitHub profile means checking whether someone’s recent public work matches your role, using signals you can judge without reading code. You are not grading syntax. You are looking for fit, activity and collaboration.

    Signals worth trusting

    • Recent, relevant activity. Work in your stack within the past year.
    • Merged pull requests to other people’s projects. It shows they can work inside a team’s standards.
    • Clear README files. Good documentation often tracks good communication.
    • Tone in issues and reviews. Read a few comments. Are they patient, specific and useful?
    • Ownership. Maintaining a project over time beats ten abandoned weekend experiments.

    Signals that mislead

    • The contribution graph. GitHub says that by default the graph shows only public-repository activity, and commits count only if the author email is connected to the account. A quiet graph can mean private work.
    • Star counts. A starred repo can be a tutorial clone or a viral side project.
    • Forks. A forked repo with no changes is just a bookmark.
    • Commit volume. Quantity says nothing about quality or difficulty.

    A two-minute profile checklist

    1. Read the bio, location and linked site.
    2. Check pinned repositories. They show what the person wants you to see.
    3. Open one pinned repo and read the README.
    4. Scan recent activity for your target stack.
    5. Check for contributions to other projects, not only their own.
    6. Decide: reach out, save for later, or skip.

    If you cannot judge technical depth, bring in an engineer for ten minutes on your top five. Our guide to reducing time to hire for technical roles shows where that review fits in the wider process.

    Step 4: Find a legitimate way to reach out

    The cleanest way to contact a GitHub developer is through details they chose to make public, such as a profile email, a personal website, or a linked social account. Do not guess, scrape or reverse-engineer private addresses.

    Be careful with commit emails. GitHub lets users keep their email private and assigns a noreply address for web-based operations, so many commits will show no usable address. Older commits may still carry a real address, and that does not mean the person wants recruiter email there.

    Better options, in order:

    1. A public email on the profile.
    2. A contact link on their personal site or blog.
    3. A public social handle listed on the profile.
    4. A thoughtful comment on a project, only when it is relevant and non-promotional.

    That last one deserves a warning. Use it rarely. GitHub’s policies restrict advertising in other users’ accounts, as covered below.

    Sourcing on GitHub is allowed, but GitHub’s Acceptable Use Policies restrict how you use its data, and privacy law adds duties once you hold a person’s details. This is not legal advice. Check current terms and ask counsel for your situation.

    What GitHub’s policies say

    The GitHub Acceptable Use Policies include these points relevant to recruiters:

    • Scraping. The policy defines scraping as extracting information via an automated process such as a bot or webcrawler, and it limits what you may do with scraped data. Use the official search and API instead, and respect rate limits. GitHub’s docs list 30 authenticated search requests per minute for most endpoints.
    • Spam. You may not use information from the service for spamming purposes, including sending unsolicited emails to users or selling personal information.
    • Privacy. Personal information you collect may only be used for the purpose the user authorized, in line with GitHub’s Privacy Statement.
    • Advertising. You may not advertise in other users’ accounts, such as by posting excessive bulk content in issues.

    The practical reading: personal, relevant, low-volume outreach sits on the safer side. Mass-mailing a harvested list is the behavior the policy targets. Read the full policy yourself, because wording changes.

    GDPR and privacy law

    If you source candidates in the EU or UK, privacy law applies even to public data. Article 14 of the GDPR covers personal data not obtained from the person directly. It requires you to tell them who you are, why you hold their data, the legal basis and their rights. You must do this within a reasonable period and at the latest within one month, or at the time of first communication if you contact them.

    That has simple workflow consequences:

    • Put a short privacy notice or link in your first message.
    • Collect only what you need for the role.
    • Delete data on request and when it is no longer needed.
    • Record where each candidate came from.

    Ethics beyond the law

    Developers talk. A bad recruiter message gets screenshotted. Do not use someone’s open-source work as a hook if you did not read it. Do not reference private details. Do not email people ten times.

    Step 5: Outreach angles that get replies

    The best GitHub outreach cites one specific piece of work and explains why it matters to the role. Generic flattery gets deleted.

    Four angles that work

    1. The specific contribution. “Your pull request adding retry logic to the queue client is close to a problem our team is solving.”
    2. The technical problem. Lead with the hard problem, then the company.
    3. The stack match. Name the exact tools they use, not a keyword list.
    4. The honest context. State level, location, pay range if allowed, and remote policy upfront.

    Message rules

    • Keep it under 120 words.
    • Mention one real detail from their work.
    • Ask a small question, not for a call.
    • Offer an easy opt-out.
    • Follow up once, then stop.

    For more on getting replies, see our breakdown of LinkedIn Recruiter alternatives in 2026, which compares channels beyond InMail.

    Where AI sourcing helps

    AI sourcing tools search large public-profile datasets with natural-language queries, then rank matches so you review fewer, better candidates. They help most where manual GitHub work is slowest.

    Task Manual on GitHub Where AI helps
    Finding candidates Stack qualifiers, test variants, repeat Describe the role in plain language and get ranked matches
    Profile review Open dozens of tabs Summaries of relevant work
    Contact details Hunt through sites and profiles Enrichment of public data
    Outreach Copy, paste, track Sequenced follow-ups with tracking
    Judging quality Needs human judgment Still needs you

    ConnectDevs approaches this in three parts. AI candidate sourcing (Scout) runs natural-language search over a published graph of 800M+ public profiles and returns ranked matches with reasoning. Candidate enrichment fills in contact and profile data. Pilot runs multi-touch outreach from your own Gmail. None of it replaces an ATS, and none of it replaces reading the candidate’s work.

    For a realistic view of what automation covers, read what AI recruiting agents actually automate, and what still needs a human. If you are comparing tools, our SeekOut alternatives in 2026 roundup is a good next stop.

    FAQ

    Is it legal to source developers on GitHub?

    Yes, viewing public profiles is allowed. The limits are on how you use the data. GitHub’s Acceptable Use Policies bar spamming and unsolicited bulk email using its data, and GDPR applies if you hold EU or UK personal data. Ask counsel about your case.

    Can I scrape GitHub for developer emails?

    You should not. GitHub’s policies restrict using information from the service for spam and limit scraping. Use the official search, respect rate limits, and only contact people through details they made public.

    How do I find developers on GitHub by location?

    Search with the location: qualifier, for example language:python location:london type:user. Location is free text, so test spelling variants and nearby cities.

    How do I judge a developer on GitHub if I cannot code?

    Check recent activity in your stack, merged pull requests to other projects, README quality and tone in issues. Ignore stars and commit counts. Ask an engineer to review your shortlist.

    Does a quiet GitHub profile mean a weak developer?

    No. The contribution graph shows only public-repository activity by default, and many engineers work in private repos. Treat a quiet profile as unknown, not negative.

    Do I need a paid tool to source developers on GitHub?

    No. Free GitHub search handles everything in this guide. Paid and AI tools save time at scale, but the method works without them.

    Conclusion

    To source developers on GitHub well, search repositories first, filter people with qualifiers, read recent collaborative work, and reach out through public channels with one specific, honest message. Stay inside GitHub’s policies and privacy law, and keep volume low.

    The manual method works, but it is slow. That is where AI sourcing earns its place: discovery and ranking at speed, with your judgment on the final call. To see how this fits a fuller process, read our guide to screening high-volume job applicants next.

    Ready to find developers faster? Start sourcing candidates with AI candidate sourcing and describe your ideal engineer in plain language.

  • Boolean Search for Recruiters: Strings, Examples, and When to Switch to AI

    Boolean Search for Recruiters: Strings, Examples, and When to Switch to AI

    Boolean search for recruiters is still the fastest way to turn a vague job brief into a precise list of profiles, as long as you know the operators, the traps, and where the method stops working. This guide covers the syntax, ready-to-use strings for common engineering and non-technical roles, and a clear test for when to move part of your sourcing to natural-language AI search.

    TL;DR

    • Five operators do almost all the work: AND, OR, NOT, quotation marks, and parentheses. On LinkedIn the operators must be uppercase.
    • Build every string in three blocks: title synonyms, must-have skills, exclusions. Wrap each block in parentheses.
    • Boolean can only find the words you predicted. Candidates who describe their work differently stay invisible, and you never see them.
    • Use Boolean when the role has stable, standard vocabulary. Use natural-language search when the skills are new, niche, or described in many different ways.
    • Test any string on a small result set first. If the top 20 profiles are mostly wrong, fix the string before you send a single message.

    Table of contents

    Boolean search is a way of combining keywords with logical operators so a search engine returns only the profiles that satisfy your conditions. Recruiters use it in LinkedIn, in applicant databases, and in Google to find candidates by job title, skill, location, or employer. The name comes from Boolean logic, where every condition is either true or false.

    A plain keyword search for python developer returns everyone who mentions either word anywhere. A Boolean string such as ("python developer" OR "python engineer") AND (Django OR FastAPI) NOT intern returns only people matching the exact phrases you chose, with the exclusions applied.

    The five operators you actually need

    Most of the long cheat sheets online list a dozen operators. Five cover nearly every real search, and these are the ones LinkedIn’s search box reliably understands.

    Operator What it does Example
    AND Requires both terms "data engineer" AND Spark
    OR Accepts either term, used for synonyms "site reliability engineer" OR SRE
    NOT Removes profiles containing a term "backend developer" NOT intern
    ” “ Matches an exact phrase "machine learning engineer"
    ( ) Groups terms so the logic runs in the order you intend (Python OR Go) AND Kubernetes

    Two syntax rules cause most failed searches. First, AND, OR, and NOT generally need to be uppercase on LinkedIn, otherwise they are read as ordinary words. Second, wildcards such as develop* are not supported in LinkedIn search, so list the variants you want (developer OR development) instead. LinkedIn also changes search behavior by plan and product, so confirm the details in your own account’s help documentation before you build a saved search around an edge case.

    How to build a Boolean string in three blocks

    A string that grows one keyword at a time becomes unreadable fast. Build it in three blocks, then combine them.

    1. Titles block. List every realistic title for the role inside one set of parentheses, joined with OR. For a platform role: ("platform engineer" OR "infrastructure engineer" OR "devops engineer" OR "site reliability engineer" OR SRE).
    2. Skills block. List the one or two must-have skills, each with its common variants: (Kubernetes OR K8s) AND (Terraform OR Pulumi). Keep this block short. Every extra AND removes people who may be a strong fit but phrase things differently.
    3. Exclusions block. Add NOT terms only for noise you have actually seen in the results: NOT (intern OR student OR recruiter).

    Combine them with AND between blocks. Then run the search and read the first 20 results. If more than a quarter are clearly wrong, the usual cause is a title that is too generic or a missing exclusion. Change one block at a time so you know what fixed it.

    Copy-paste Boolean strings

    These are starting points. Adjust the synonyms to the stack and seniority your hiring manager describes.

    Backend engineer

    ("backend engineer" OR "backend developer" OR "software engineer")
AND (Go OR Golang OR Java OR Python)
AND (microservices OR "distributed systems")
NOT (intern OR student OR junior)

    Platform or DevOps engineer

    ("platform engineer" OR "devops engineer" OR "site reliability engineer" OR SRE)
AND (Kubernetes OR K8s)
AND (Terraform OR Pulumi OR CloudFormation)
NOT (intern OR student)

    Machine learning engineer

    ("machine learning engineer" OR "ML engineer" OR "applied scientist")
AND (PyTorch OR TensorFlow)
AND (deployment OR MLOps OR "model serving")
NOT (intern OR student)

    Sales development or account executive

    ("account executive" OR "sales executive" OR "enterprise sales")
AND (SaaS OR B2B)
AND (quota OR pipeline OR "closed won")
NOT (intern OR assistant)

    Technical recruiter (for agencies hiring their own)

    ("technical recruiter" OR "tech recruiter" OR "technical sourcer")
AND (engineering OR software OR "IT staffing")
NOT (coordinator OR intern)

    Notice that none of these strings tries to be clever. They use plain titles, two or three skills, and a short exclusion list. Strings with fifteen nested conditions usually return fewer candidates, not better ones.

    Google X-ray search for profiles outside your tool

    X-ray search uses Google’s site: operator to search a specific website’s public pages, which lets you look for candidates on sites where you have no recruiter license. The pattern is site:[domain] "job title" "skill" "location". For example:

    site:github.com "machine learning" (Python OR PyTorch) "Berlin"

    Google uses a minus sign for exclusions, so write -jobs -careers rather than NOT. X-ray results are public pages only, so treat them as leads and verify details before outreach. For more places to look beyond a single network, see our guide to finding passive candidates beyond LinkedIn.

    Common Boolean mistakes

    • Missing parentheses. Python OR Java AND AWS does not mean what most people think it means. Group the OR terms: (Python OR Java) AND AWS.
    • Too many AND conditions. Each one shrinks the pool. Start with two skills, not six.
    • Searching only the title you were given. The hiring manager’s job title is rarely the title candidates use. Include at least four synonyms.
    • Copying strings from the internet unchanged. A string tuned for another company’s stack will return their candidates, not yours.
    • Never testing exclusions. A careless NOT can remove the very people you want, for instance excluding “manager” while hiring an engineering manager.

    Where Boolean search breaks down

    The core limit is not syntax. It is vocabulary. A Boolean string can only match words the recruiter predicted in advance. As one sourcing vendor puts it, a keyword search only reaches the vocabulary you predicted, and strings cannot show you who they missed, only who they matched (Metaview). An infrastructure engineer who writes “self-managed clusters” on their profile instead of “Kubernetes” will never appear for a string built around Kubernetes.

    That matters most in four situations:

    1. New or fast-changing skills. Titles and terms for AI and platform roles shift every few months, so strings go stale.
    2. Niche combinations. “Someone who has scaled a payments system in a regulated market” is a concept, not a keyword.
    3. Non-standard careers. Founders, freelancers, and people who moved between industries describe their work in their own words.
    4. Volume. Writing, testing, and maintaining strings for ten open roles is real recruiter time that does not go into talking to candidates.

    A caveat applies to the AI side too. Vendor benchmark claims for AI sourcing are largely vendor-run, and even Metaview acknowledges nobody has yet measured whether AI-surfaced candidates are better hires than those a skilled sourcer would have found. Judge any tool, including ours, on the quality of the shortlist it gives you on your own roles.

    Boolean vs. natural-language search: how to choose

    Natural-language search lets you describe the person you want in plain sentences, and a semantic model finds profiles that match the meaning rather than the exact words. For a deeper look at why meaning-based matching differs from keyword matching, read our piece on intent-based matching vs. keyword search.

    Situation Better fit Why
    Standard title, stable skills (for example, Java developer) Boolean Vocabulary is predictable, results are fast and controllable
    Niche or emerging skills Natural language Candidates describe the work in varied terms
    Brief is a story, not a keyword list Natural language Concepts and context do not reduce to operators
    You need an auditable, repeatable search Boolean The logic is explicit and can be saved and shared
    Many open roles, small team Natural language Less time spent writing and tuning strings

    Most strong sourcers end up using both. They run a tight Boolean search for the obvious pool, then run a natural-language search on the same brief to find the candidates the string missed. Compare the two lists. The profiles that appear only in the second list tell you which synonyms your string was blind to, and you can add them back.

    ConnectDevs Scout is built for the second half of that workflow. You describe the role in plain language, and Scout’s AI candidate sourcing converts it into a structured search over a public-profile graph, ranks the results, and shows why each person matched so you can check the reasoning instead of trusting a score. It supports recruiter judgment and does not replace it, and it does not guarantee a candidate’s availability or interest.

    Whichever method finds the profile, the next problem is missing contact details and stale data. That is where candidate enrichment fits in, and if you are weighing tools more broadly, our comparison of LinkedIn Recruiter alternatives covers pricing and trade-offs.

    FAQ

    What is Boolean search for recruiters?

    Boolean search for recruiters is a method of combining keywords with the operators AND, OR, NOT, quotation marks, and parentheses to filter candidate profiles precisely. It lets you specify titles, skills, and exclusions in a single query instead of scanning results by hand.

    Does LinkedIn support Boolean search?

    Yes. LinkedIn supports AND, OR, NOT, quotation marks, and parentheses in keyword search, and operators should be typed in uppercase. Wildcards are not supported. Exact limits and field behavior vary by plan, so check the help documentation for your LinkedIn product.

    What is an X-ray search?

    An X-ray search uses Google’s site: operator to search the public pages of a single website, such as GitHub or a company team page, for candidate information. It is useful for finding leads outside a paid recruiting tool, though results are limited to public pages.

    Is Boolean search still worth learning in 2026?

    Yes. It is fast, transparent, and effective for roles with standard titles and skills, and the logic carries over to applicant databases and ATS filters. It works best alongside natural-language search, which covers the vocabulary Boolean cannot predict.

    How long should a Boolean string be?

    As short as it can be while still returning relevant profiles. Two or three skills plus a title block and a short exclusion list is usually enough. If you keep adding conditions to fix results, the problem is probably the title synonyms, not the skill list.

    What is the difference between Boolean and natural-language candidate search?

    Boolean matches the exact words and logic you specify. Natural-language search interprets a plain-English description and matches on meaning, so it can find profiles that use different wording. Boolean is more controllable; natural language is broader.

    Conclusion

    Learn the five operators, build strings in three blocks, and test them on a small sample before you commit to outreach. Keep Boolean for roles with predictable vocabulary, and add natural-language search wherever your strings keep missing people. If you want to see how a plain-language brief performs against your current strings on a live role, book a demo and bring a brief you have already sourced for.

  • Boolean Search vs Natural Language Search: How Recruiters Should Source in 2026

    Boolean Search vs Natural Language Search: How Recruiters Should Source in 2026

    Boolean Search vs Natural Language Search: How Recruiters Should Source in 2026

    Most recruiters learned Boolean search the hard way: a string that returns 40,000 profiles, then a tighter one that returns six. The Boolean search vs natural language search debate is now practical, not theoretical, because LinkedIn and most AI sourcing tools let you type a plain-English description of the person you want. This guide shows where Boolean still wins, where natural language wins, and a workflow that uses both, with real examples for technical roles.

    TL;DR

    • Boolean search matches exact keywords and operators. It is precise and predictable, but it misses candidates who describe themselves differently.
    • Natural language search interprets intent, related titles, and adjacent skills. It finds more relevant people faster, but you must check why each result matched.
    • Use Boolean when you need exact, auditable filters. Use natural language to explore and to find people with unusual titles.
    • The best workflow starts with a plain-language search, reads the results, then adds Boolean constraints only where precision matters.

    Table of Contents

    What Is Boolean Search in Recruiting?

    Boolean search is a way of searching by combining keywords with logical operators such as AND, OR, and NOT, plus quotation marks for exact phrases and parentheses for grouping. A recruiter writes the rules, and the database returns only profiles that satisfy them literally.

    That literal behavior is both its strength and its weakness. If you write ("Software Engineer" OR "Software Developer") AND (Go OR Golang), you get profiles containing those words. You do not get the engineer whose title is “Member of Technical Staff” or “Backend Developer” unless you thought to include them. Good Boolean work is mostly the craft of anticipating every way a candidate might describe themselves.

    What Is Natural Language Candidate Search?

    Natural language candidate search lets a recruiter describe the person they want in plain English, such as “backend engineer with production Go experience at an early-stage startup, three to five years in.” The system interprets the request, maps it to related titles, skills, seniority, and career context, and returns ranked results.

    The idea is no longer niche. LinkedIn announced AI-assisted search for Recruiter in October 2023, letting recruiters describe a hiring need and receive recommendations from a broader pool. HR Dive reported that early testing showed 74% of recruiters saw time savings. That is a vendor-reported early-testing figure, so read it as a signal of direction rather than a benchmark. Sourcing platforms such as ConnectDevs work the same way: AI candidate sourcing converts role requirements into a structured search and shows the reasoning behind each result for recruiter review.

    Side-by-Side Comparison

    Factor Boolean search Natural language search
    Input Keywords and operators Plain-English description
    Matching Exact terms only Related titles, skills, and context
    Learning curve Steep; skill takes practice Low; usable by hiring managers and founders
    Predictability High; same string, same rules Lower; ranking depends on the model
    Auditability Easy to explain exactly why a profile matched Depends on whether the tool shows its reasoning
    Risk Missing good candidates with different wording Including loosely relevant candidates
    Best for Narrow, well-defined searches and compliance-sensitive filters Exploring a market and roles with inconsistent titles

    Examples for a Backend Engineer Search

    Boolean version

    ("Backend Engineer" OR "Software Engineer" OR "Platform Engineer") AND (Go OR Golang) AND ("distributed systems" OR microservices) NOT (recruiter OR sourcer)

    This works if candidates use those exact words. It excludes anyone whose profile says “Staff Engineer, payments infrastructure” without the phrase “distributed systems.”

    Natural language version

    Backend engineer who has shipped production Go services at a startup or scale-up, comfortable with distributed systems, three to six years of experience, based in Europe.

    A semantic system can surface the payments infrastructure engineer above because it understands the skills overlap. You still need to open profiles and confirm the evidence.

    Combined version

    Run the plain-language search first. Scan the top results and note the job titles and skill words that appear. Then add a Boolean filter only for hard requirements, such as a work-authorization region or a specific technology you cannot compromise on.

    A Workflow That Uses Both

    1. Write the requirement as a paragraph. Include the outcome the person will own, not only a title.
    2. Run a natural language search and review the first 20 to 30 results for relevance.
    3. Harvest vocabulary. Note titles and skill terms strong candidates use. This is also how you improve future Boolean strings.
    4. Add hard filters with Boolean or structured filters for location, seniority, and must-have skills.
    5. Check the evidence. Confirm the reason each candidate matched before you send outreach.
    6. Enrich the shortlist so work history and contact routes are complete. Our guide to candidate enrichment explains why this step saves outreach time.

    For more ways to reach engineers who never update a profile, see our tactics for finding passive candidates beyond LinkedIn, and read why intent-based matching outperforms keyword search for the reasoning behind the shift.

    Mistakes to Avoid

    • Trusting the ranking blindly. A ranked list is a starting point. Semantic tools can rank a loosely related profile highly.
    • Over-specifying. Long plain-language prompts stuffed with every preference can narrow results the same way a bad Boolean string does.
    • Dropping Boolean entirely. For regulated or audited searches, exact filters are easier to document and defend.
    • Ignoring bias. Any filter, keyword or semantic, can screen out people unfairly. Review the composition of results and keep a human in the decision.

    FAQ

    What is the difference between Boolean search and natural language search?

    Boolean search matches the exact keywords and operators a recruiter writes. Natural language search accepts a plain-English description and interprets meaning, including related titles and skills, to return ranked candidates.

    Is Boolean search still worth learning in 2026?

    Yes. It gives precise, explainable filters, works in nearly every database, and helps you check what an AI tool returned. It is now one tool among several, not the only way to search.

    Is natural language search more accurate than Boolean?

    Not automatically. It usually finds more relevant people with unusual titles, but Boolean is more predictable for narrow searches. Accuracy depends on the tool and on how well you review the results.

    Can natural language search replace recruiters?

    No. It speeds up discovery and research. Judging fit, motivation, and context, and making the hiring decision, remain human responsibilities.

    Does LinkedIn Recruiter support natural language search?

    LinkedIn introduced AI-assisted search in Recruiter in 2023 and has continued to add AI features. Availability depends on your license, so check your current plan. See our list of sourcing alternatives if you want to compare options.

    How do I write a good natural language search for engineers?

    Describe the work the person will do, the stack, the seniority, and the type of company they should have worked in. Avoid listing every nice-to-have. Start broad, review results, then narrow.

    Conclusion

    Treat Boolean and natural language search as complementary. Start with intent to discover people you would not have found with keywords, use what you learn to sharpen your filters, and verify every shortlisted profile. The recruiters getting the best results are not choosing a side; they are using each method for the part of the search it does best.

    Want to see a plain-English search across an 800M+ public-profile graph? Start sourcing candidates with ConnectDevs, or start hiring free and run your next search today.

  • How to Find Passive Candidates Who Aren’t Job Hunting: 7 Tactics Beyond LinkedIn

    How to Find Passive Candidates Who Aren’t Job Hunting: 7 Tactics Beyond LinkedIn

    Roughly two out of three people in the workforce aren’t looking for a job right now. That doesn’t mean they aren’t open to the right one. LinkedIn’s own data puts the share of workers actively job hunting at just 36%, which means the majority of the talent you need is sitting outside any “open to work” filter. Learning how to find passive candidates is the actual job of sourcing, since anyone can message people who already raised their hand.

    The problem is that most recruiters’ idea of “passive sourcing” still means one thing: search LinkedIn, filter by title and years of experience, send an InMail, wait. That approach is getting worse, not better. InMail credits are limited, and the same 200 senior engineers in your metro area get the same templated message from twelve other recruiters this week.

    This guide skips the LinkedIn-alternative debate (we’ve covered why LinkedIn Recruiter is failing recruiters and the pricing problem pushing teams to diversify elsewhere) and gets straight to seven sourcing channels and techniques you can use this week to reach people who aren’t scrolling job boards.

    TL;DR

    • About 64% of the workforce is passive, and passive candidates make up 69% of people who actually reply to well-targeted recruiter outreach. They’re not unreachable, just poorly targeted.
    • For technical roles, GitHub, Stack Overflow, and open-source contribution history surface real skill signal that a resume or LinkedIn headline can’t.
    • Niche Slack and Discord communities, conference speaker lists, and alumni networks put you in front of engaged practitioners before they’re job hunting.
    • Referral and alumni network mining — done systematically, not with a Slack message asking “know anyone?”, is still one of the highest-converting channels available.
    • AI-powered natural-language search lets you run all of these channels at once instead of manually checking six tabs, which is where tools like ConnectDevs fit in.

    Why passive candidates are worth the extra effort

    Passive candidates aren’t harder to hire because they’re uninterested. They’re harder to hire because they’re not looking where recruiters look. They’re not refreshing job boards, not updating an “open to work” badge, and definitely not opening a fifth InMail this month from a stranger.

    But the data on what happens after you reach them is encouraging. In a 2026 analysis of over 400,000 candidate replies, Pin found that 69% of people who responded to recruiter outreach were not actively job hunting. Most described themselves as either “not looking” or “passive but open.” The same research found that candidates who’d recently changed roles or updated their profile replied at meaningfully higher rates, meaning timing and relevance matter more than volume.

    Compare that to raw InMail performance: average response rates across industries generally fall in the 18-25% range, with talent acquisition specifically closer to 12%. Passive candidates respond, but only when the message and the channel are right. That’s the whole game.

    1. Mine GitHub and Stack Overflow for real skill signal, not job titles

    For engineering roles, a LinkedIn title tells you what a company called someone. A GitHub profile tells you what they can actually do. Commit history, repo README files, issue threads, and code review comments show depth of skill, communication style, and current interests — all things a resume filters out.

    Recruiter scenario: You need a senior Rust engineer with distributed systems experience. Instead of searching LinkedIn for “Rust Engineer,” search GitHub for repos tagged with relevant frameworks (e.g., tokio, actix), sort contributors by recent commit activity, and look at who’s answering hard architecture questions on Stack Overflow’s Rust tag. You’ll surface people actively building in the language today, a much stronger signal than a five-year-old job title.

    • Use GitHub’s advanced search to filter by language, followers, and recent activity.
    • Check who maintains or frequently contributes to libraries your own stack depends on.
    • On Stack Overflow, sort by reputation earned in the last 30-90 days, not lifetime reputation, to find people currently active.

    2. Go where practitioners already gather: niche Slack and Discord communities

    Every technical discipline has a handful of Slack workspaces or Discord servers where the real conversation happens, far more candid and current than anything posted publicly. Data engineering has Locally Optimistic and dbt Slack. DevRel has multiple invite-only Slacks. Design has communities built around specific tools. Product has Mind the Product and similar groups.

    Recruiter scenario: A hiring manager wants a platform engineer who’s used Kubernetes at scale. Rather than another LinkedIn search, you join two or three Kubernetes-focused Slack communities, read the #careers or #jobs channel to see who’s helping others (a strong signal of seniority), and start genuine conversations in threads before ever mentioning a role.

    • Join as a participant first, not a recruiter with a pitch. Most communities ban cold recruiting in public channels.
    • Look for people who consistently answer other members’ technical questions; they tend to be senior and well-regarded by peers.

    3. Turn open-source contribution graphs into a pre-vetted shortlist

    Open-source activity is one of the few sourcing signals that’s both public and hard to fake. Someone with merged pull requests to a widely-used project has already been code-reviewed by strangers with no incentive to be generous.

    Recruiter scenario: You’re hiring for a company that runs on a specific open-source framework (say, a popular JavaScript testing library). Pull up the project’s GitHub “Insights” tab, look at top contributors by commits and PRs merged in the last six months, and cross-reference their public profiles for location and current employer. This surfaces people who understand the exact tooling your team uses, often before that skill shows up on any resume.

    • Prioritize contributors with merged (not just opened) pull requests. Merged code reflects accepted quality, not just activity.
    • Track contributors to adjacent projects in the same ecosystem, since skills often transfer across related tools.

    4. Turn conference speaker and attendee lists into a sourcing list

    Conference speakers have self-selected as people confident enough in their expertise to present it publicly, and most conference sites publish full speaker bios, talk abstracts, and sometimes attendee lists or Slack/Discord communities tied to the event.

    Recruiter scenario: You’re sourcing for a data science lead. Pull the speaker list from two or three recent data science conferences (PyData, relevant regional meetups, vendor-hosted summits), read talk abstracts to find people speaking on topics that map to your role, and reach out referencing the specific talk. “I watched your talk on feature store architecture” is a dramatically better opener than a templated InMail.

    • Check conference apps and Sched/Whova-style attendee directories, which often list titles and companies even for people who didn’t speak.
    • Meetup.com groups tied to a technology are a lower-effort version of the same tactic. Organizers and regular attendees are usually named publicly.

    5. Mine your best employees’ and alumni networks systematically

    Most referral programs stop at “let us know if you know anyone.” That leaves the actual mining work undone. Your best engineers’ GitHub followers, their conference co-speakers, and their former teammates at previous companies are a far richer pool than whatever names happen to come to mind when someone gets a Slack ping.

    Recruiter scenario: Instead of a generic referral ask, sit with a top performer and go through their LinkedIn connections or GitHub follower list together, asking specifically: “Who on this list would you want to work with again?” Do the same with company alumni groups. Many companies have unofficial or official alumni Slack/LinkedIn groups where former employees stay loosely connected, and people who left on good terms are often warm to hearing from a former colleague turned recruiter.

    • Search LinkedIn or company alumni pages for “worked at [target company]” plus your target skill for a fast way to find people with a specific employer pedigree.
    • University alumni associations and alumni-only LinkedIn groups work the same way for early-career or specialized-degree roles.
    • This kind of systematic network mining is really a talent-mapping exercise: mapping who exists and where before you ever reach out. If your team is doing this ad hoc, it’s worth reading how talent mapping, sourcing, and pipelining differ as distinct disciplines so the effort compounds instead of resetting every req.

    6. Watch the conversation on X/Twitter and niche industry newsletters

    Technical and specialist communities still have an active presence on X, often organized around hashtags, subject-matter Twitter lists, or reply threads under well-known accounts in a field. Industry newsletters (think Console.dev for engineering, Lenny’s Newsletter for product, or smaller vertical-specific ones) also have comment sections and referenced communities worth mining.

    Recruiter scenario: You’re hiring a machine learning engineer with LLM evaluation experience. Search X for recent threads on evaluation frameworks, note who’s contributing substantive technical replies (not just retweets), and check their bio for current role. Subscribe to two or three newsletters in the space and track who shows up as a quoted expert or guest contributor over a few issues.

    • Twitter/X “Lists” curated by well-known people in a field are a fast way to find who the community considers credible.
    • Treat this channel as an ongoing watchlist rather than a one-time search. It rewards patience more than the others.

    7. Use AI-powered natural-language search to run all six channels at once

    Manually checking GitHub, three Slack communities, a conference speaker list, alumni groups, and X threads for every open req doesn’t scale past one or two roles at a time. This is the part of passive sourcing that’s genuinely changed in the last two years: natural-language, AI-driven search can now query across public profiles, code repositories, and professional activity signals the same way you’d describe a candidate to a colleague — “senior backend engineer, contributed to open-source Go projects, based in Austin, not actively applying anywhere”, instead of building a boolean string for one platform at a time.

    Recruiter scenario: Instead of manually working through tactics one through six for every requisition, you describe the role and ideal background in plain language to an AI candidate sourcing tool, and it pulls and ranks candidates across the open web, surfacing enrichment signals like open-source activity and community involvement alongside standard profile data, so your team spends its time on outreach and conversations, not tab-switching.

    • This is the core idea behind what makes a hiring platform AI-native rather than a search box bolted onto an old database: it’s built to search intent and signal, not just keywords.
    • It also connects directly to why intent-based matching outperforms keyword search, since passive candidates rarely use the exact job title you’re searching for, so keyword matching misses them by design.
    • The enrichment layer matters as much as the search. AI candidate enrichment is what turns a bare GitHub username or conference bio into a complete, verifiable profile you can actually act on.

    Quick channel comparison

    Channel Best for Effort level Signal strength
    GitHub / Stack Overflow Technical/engineering roles Medium Very high
    Slack / Discord communities Any specialist discipline High (relationship-based) High
    Open-source contributions Engineering, specific frameworks Medium Very high
    Conference speakers/attendees Senior/expert-level roles Low-medium High
    Referral & alumni mining Any role, culture fit Medium Very high
    X/Twitter & newsletters Thought-leader and niche roles Low, but slow Medium
    AI natural-language search Scaling across all of the above Low High

    Putting it together: a realistic weekly sourcing mix

    No single channel above replaces the others. They compound. A recruiter working a senior backend role might spend Monday running an AI-powered natural-language search to build an initial long list, Tuesday cross-referencing GitHub contribution activity for the strongest technical signal, and the rest of the week on warm outreach through alumni connections and community engagement for the top 15-20 names. The point is to stop treating LinkedIn search as the only tool in the kit.

    If your process for tracking who you’ve mapped, sourced, and pipelined across these channels is a spreadsheet, that’s usually the first thing worth fixing. A talent intelligence platform built for how sourcing actually works today keeps that signal in one place instead of scattered across six browser tabs.

    FAQ

    What percentage of candidates are passive versus actively job hunting?

    Roughly 64% of the workforce is passive at any given time, based on LinkedIn’s own data showing about 36% of workers are actively looking for a new role. That means the majority of qualified people for almost any role aren’t applying anywhere. They have to be found and reached directly.

    Is it worth the effort to source passive candidates instead of just posting a job?

    Yes, especially for senior or specialized roles where the strongest people are rarely browsing job boards. Research on candidate replies found that 69% of people who respond positively to recruiter outreach weren’t actively job hunting. The effort pays off when the message is targeted.

    How do I message a passive candidate without sounding like spam?

    Reference something specific and real: a conference talk they gave, a pull request they merged, a comment they made in a community you’re both part of. Generic “exciting opportunity” openers get ignored regardless of channel. Specificity is what separates a message that gets read from one that gets reported.

    Do these tactics work for non-technical roles, or just engineering?

    The channels shift, but the underlying logic works everywhere. Designers have Behance and Dribbble communities, product managers have Mind the Product and similar Slack groups, marketers have niche newsletters and X communities. The approach is the same: find where practitioners gather and demonstrate expertise publicly, then engage there before pitching a role.

    How many sourcing channels should I actually use for one open role?

    Two or three well-matched channels usually beat spreading thin across all seven. For a technical role, GitHub plus a relevant Slack community plus referral mining is often enough. Add conference or newsletter tracking for senior, hard-to-fill, or leadership-level searches where the pool is smaller and more visible.

    How is AI-powered sourcing different from just automating LinkedIn searches?

    Automating a LinkedIn search still only searches LinkedIn, and still relies on candidates using the keywords you searched for. AI-driven natural-language sourcing searches across the open web, including code repositories, professional activity, and public profiles, and matches on described intent and signal rather than exact keyword overlap, which is closer to how a good sourcer actually thinks about a role.

    What’s the single highest-converting tactic in this list?

    Referral and alumni network mining consistently converts best because there’s already a trust relationship, even a loose one. It’s also the most time-intensive to do well, which is exactly why it’s usually under-invested. Most teams ask for referrals once and stop rather than mining the network systematically.

    The bottom line

    Finding passive candidates isn’t about one clever trick. It’s about showing up where the people you want already spend their time, with something more specific to say than a templated InMail. GitHub activity, community involvement, conference presence, and network mining all surface people before they’re job hunting, and each one gets stronger when it’s not done from scratch for every single req.

    The hard part isn’t identifying these channels — it’s running them consistently without burning your whole week on tab-switching. That’s the gap AI candidate sourcing is built to close. Explore AI candidate sourcing with ConnectDevs to see how natural-language search, enrichment, and matching work together across the channels above.