AI Recruiting Academy › Module 2: Finding candidates with AI
The short answer
Boolean search finds profiles that contain the exact words you type, joined with AND, OR and NOT. Semantic search finds profiles that mean the same thing as your request, even when the words are different, so “people leader” can surface someone who wrote “managed a team of twelve engineers”.
Semantic search finds more of the right people. Boolean search is precise, predictable and easy to check. The best sourcers now use both: describe the role in plain English to explore, then use a Boolean string to control and check the results.
What is Boolean search?
Boolean search is a way of writing a search with logic words. AND means both terms must appear. OR means either term can appear. NOT removes profiles with a term. Quotation marks keep a phrase together and brackets group the OR terms.
A string for a payroll role might look like this:
("payroll specialist" OR "payroll administrator") AND (ADP OR Paylocity) AND "multi state" NOT intern
Boolean has been the core sourcing skill for decades because it works almost everywhere: LinkedIn, Google searches of sites like GitHub, job board resume databases and most applicant tracking systems. Our guide to Boolean searches covers the basics, and LinkedIn Boolean search covers the operators LinkedIn supports.
What is semantic search?
Semantic search matches meaning instead of matching words. The software turns each profile, and your request, into a long list of numbers that describes what the text is about. These lists are called embeddings. Profiles whose numbers sit close to your request’s numbers are treated as good matches, whatever words they use.
That is why semantic search can connect “talent acquisition partner” with “recruiter”, or “built our payroll process from scratch” with “payroll implementation”. A keyword search would miss both unless you had thought to type every variation.
LinkedIn explained the problem it was solving in its engineering write up on Hiring Assistant: keyword matching fails because the same qualification can be described in dozens of ways, and its Boolean search strategy returned zero results for nearly half of the queries it was tested on.
How do semantic and Boolean search compare?
| Boolean search | Semantic search | |
|---|---|---|
| What you type | Keywords joined with operators | A plain English description |
| What it matches | The exact words | The meaning behind the words |
| Finds synonyms | Only the ones you list | Yes, on its own |
| Skills that are hard to name | Weak | Better, such as “comfortable with ambiguity” |
| Can you see why someone matched? | Yes, the words are on the profile | Only if the tool explains it |
| Same search, same results? | Yes | Not always, as models and data change |
| Common failure | Too few results, or zero | Plausible looking people who miss a must have |
What changed for recruiters?
The skill has moved from writing strings to writing briefs and checking results. You no longer need to predict every way a candidate might describe their job. You do need to say clearly what matters, and then spot when the tool has drifted.
LinkedIn Recruiter shows how the two now live side by side. Its AI Search turns a plain language request, such as “Find me Marketing Managers in Dublin with sales enablement experience”, into filters for location, skills and titles, and it still accepts Boolean and keyword searches. Advanced AI Search, available to Recruiter and RPS+ customers with English settings, looks for skills that rarely appear as keywords, such as the ability to lead large cross functional work.
Behind the scenes, LinkedIn’s agent runs Boolean, filter based and semantic search at the same time and blends the results. Boolean did not disappear. It became one input among several.
Where does each kind of search fail?
Boolean fails quietly on language. Miss one synonym and you miss everyone who uses it. Add one AND too many and you get nothing. It also cannot tell a candidate who led a payroll migration from one who sat near one.
Semantic fails quietly on precision. It can rank someone highly because their profile sounds similar, even though they lack a hard requirement such as a license or a location. It is also harder to audit, because there is no single word you can point to and say “this is why”.
So each needs the other. Semantic search widens the net. Boolean and filters enforce the must haves.
How do you use both in one search?
| Step | What you do |
|---|---|
| 1. Describe | Write the role in plain English, with must haves and nice to haves |
| 2. Inspect | Look at the filters or keywords the tool built from your request |
| 3. Enforce | Add Boolean terms or filters for anything that is truly required, such as a license |
| 4. Sample | Read the first twenty results and note what is wrong |
| 5. Refine | Rate results, adjust the brief and run it again |
| 6. Save | Keep the brief and the final string so the next search starts ahead |
Quick answers
Is Boolean search dead?
No. It is still how you search Google, many job boards and most ATS databases, and it is the fastest way to check what an AI tool is doing. What has changed is that you write fewer strings from scratch.
Can AI write Boolean strings for me?
Yes, and it is a good use of AI, but check the result. Models often use operators a platform does not support, which Lesson 2 of Module 1 covers. Our free Boolean string builder produces strings with the right syntax.
Does my ATS have semantic search?
Many now claim it. Test it: search for a skill using a word you know appears on only a few resumes, then search for a synonym. If the second search finds the same people, it is matching meaning.
Why do two AI tools return different people for the same brief?
They search different data, use different models and rank for different goals. Some favor likely responders, others favor closest fit. That is another reason to test tools on a role you have already filled.