AI Recruiting Academy › Module 2: Finding candidates with AI
The short answer
AI sourcing is software that finds and ranks potential candidates from a role description written in plain English. Instead of building a search filter by filter, you describe who you need, and the tool searches a large pool of profiles, reads them for meaning, ranks the best matches and often drafts the first message.
It makes the top of the funnel faster. It does not decide who is a good fit. The shortlist is only as good as the brief you give it, and a person still reviews every name before anyone is contacted.
How is AI sourcing different from traditional sourcing?
Traditional sourcing is manual search. You turn the job into keywords and filters, run the search, open profiles one by one and keep a list. AI sourcing moves the first three of those steps to the software, so your time goes into judging the list rather than building it.
| Step | Traditional sourcing | AI sourcing |
|---|---|---|
| Define the search | You write a Boolean string and pick filters | You write a brief in plain English |
| Find people | The platform returns exact keyword matches | The tool searches for people whose experience means the same thing |
| Sort the results | You open profiles and judge each one | The tool ranks matches and explains why each one fits |
| First contact | You write each message | The tool drafts a message for you to edit |
| Improve the search | You rewrite the string | You rate results and the tool adjusts |
How does an AI sourcing tool work?
Most tools follow the same pattern, even when the screens look different.
1. It reads your brief. A language model turns your description into structured parts: title, seniority, location and a list of qualifications.
2. It searches for meaning, not just words. The tool compares those qualifications against profiles in its database. This is semantic search, which Lesson 2 explains.
3. It ranks and explains. Candidates are scored against the brief, and good tools show which requirements each person meets and which are missing.
4. It hands you a shortlist. Many tools also find contact details and draft outreach, which Lessons 3 and 5 cover.
LinkedIn has published how its own agent does this. Its engineering team describes how Hiring Assistant searches more than a billion member profiles: a language model parses the recruiter’s free text into role details and qualifications, several search methods run side by side, and the combined results are ranked and checked by an AI evaluator before the recruiter sees them.
What results are recruiters seeing?
The published numbers are early and mostly come from vendors, so read them as direction rather than promise.
LinkedIn says its Hiring Assistant charter customers, more than 500 companies and 8,000 early users, saved over four hours per role and reviewed 62% fewer profiles, according to LinkedIn’s own analysis. In its Future of Recruiting 2025 report, 37% of organizations said they were actively integrating or experimenting with generative AI in hiring, up from 27% a year earlier, and those using it reported saving about 20% of their work week.
What those numbers do not tell you is whether the hires were better. Time saved on search is real. Quality still depends on the brief and on the recruiter reviewing the list.
Where does AI sourcing data come from?
An AI sourcer can only find people who are in the data it searches. Before you buy, ask where that data comes from, because it decides both what you will find and how much risk you carry.
| Data source | Strength | Watch out for |
|---|---|---|
| A professional network the vendor owns, such as LinkedIn | Profiles people keep up to date themselves | You can only search it inside that vendor’s product |
| A vendor database built from public profiles and other sources | Very broad reach across sites and platforms | Stale records, and data collected in ways a network’s terms forbid |
| Your own ATS and CRM | People who already know you | Old records and duplicates; Lesson 4 covers this |
The collection method matters. In 2025 LinkedIn sued the data provider Proxycurl over scraping with fake accounts. Proxycurl shut down, and the court ordered it to permanently delete the data it had collected. If a tool cannot explain where its profiles come from, treat that as a warning.
What does AI sourcing get wrong?
| Problem | What it looks like | What to do |
|---|---|---|
| Stale profiles | A strong match who changed jobs a year ago | Check the current role before you reach out |
| Cloning your last hire | Every result went to the same schools and companies | Brief on skills and outcomes, not on one example profile |
| Invented detail | A summary credits someone with experience that is not on the profile | Check every claim against the source before it reaches a candidate |
| Thin data for niche roles | A short list of weak matches for a rare specialism | Widen the location, add adjacent titles, or search by hand |
| Surprise costs | Credits used up on profiles you never contact | Check how the tool charges before running large searches |
How do you write a sourcing brief that AI can use?
The brief is the prompt, so the method from Lesson 4 of Module 1 applies. Give the tool what a good sourcer would ask the hiring manager for:
| Part of the brief | Example |
|---|---|
| Role and level | Senior payroll specialist, individual contributor |
| Must haves | Multi state US payroll, three or more years on ADP or Paylocity |
| Nice to haves | CPP certification, experience in a company of 500 to 2,000 people |
| Location | Remote, US Eastern or Central time zones |
| Deal breakers | Only bureau or outsourced payroll experience, never in house |
| Companies to skip | Current clients and your own company |
Then read the first twenty results before you trust the rest. If they are wrong, the brief is usually the problem. Tighten the must haves, rate a few results, and run it again.
Where does the recruiter stay in charge?
The rule from Lesson 3 of Module 1 holds here too: let the tool search, rank and draft, and keep a human checkpoint in front of anything that contacts a person or removes them from consideration. You approve the shortlist. You approve every message. And you own the reason someone was left off the list.
Quick answers
Will AI sourcing replace sourcers?
It replaces a lot of search work. It does not replace knowing a market, reading between the lines of a profile or persuading a passive candidate to talk. Sourcers who learn to brief and check these tools cover far more roles.
Is AI sourcing only for tech roles?
No. It works best wherever people describe their work in detail online. It is weaker for roles where candidates have thin profiles, such as many hourly and trade jobs.
Do I still need LinkedIn Recruiter?
Many teams keep it alongside an AI sourcer, because LinkedIn’s own network is the freshest data for many roles. Our guide to LinkedIn Recruiter alternatives covers when to switch, stack or stay.
How much does an AI sourcing tool cost?
Pricing ranges from monthly seats to credits per profile or contact. Each listing in the directory shows the starting price we checked on the vendor’s own site.