AI Job Matching
Matching compares two documents: what a candidate has said about themselves, and what you have said about the role. The second half is the one you control.
The short answer
- Matching compares a candidate profile and resume against a specific posting, and reports how well one answers the other.
- A vague posting produces vague matching. The posting is half the input, so job description quality is a matching input rather than a separate concern.
- A match score is an ordering that directs attention. It is not a prediction of performance and should never be the only thing a decision rests on.
- Matching surfaces candidates who fit what you wrote down. If you wrote down the wrong requirements, it will faithfully find the wrong people.
What is being compared
Two things. On the candidate side, their resume plus the structured parts of their profile: categories, skills, experience, and stated preferences such as location and work model. On your side, the posting.
The result is a relative answer, not an absolute one. There is no such thing as a good candidate score in the abstract, only a good answer to a particular posting. The same person can score strongly against one of your roles and weakly against another, and that is the system working correctly.
Your posting is half the input
This is the part employers most often miss. Matching quality is bounded by how clearly the role is described, because a vague posting gives the comparison very little to work against.
Three things in a posting move matching noticeably: describing the actual work rather than generic responsibilities, separating day-one requirements from things that can be learned, and using the vocabulary the field actually uses rather than internal terminology.
Which means job description quality is not a separate discipline from matching quality. See Creating Great Job Descriptions.
What a score means
A match score says how well a candidate answers this posting. Its purpose is ordering: it puts your attention at the useful end of a long list.
What it does not do is predict job performance. Performance depends on things no profile contains, including how someone works with your particular team and what they are like when a project goes badly. Treat a strong score as a reason to read the application properly, not as a decision.
The practical discipline is to keep reading below the top of the list. Matching optimises for what you described, and a requirement you wrote carelessly can push down a candidate you would have hired.
Where it goes wrong
It is faithful to what you wrote. If the posting lists a technology as essential that is genuinely learnable in a fortnight, matching will honour that and rank capable people below less capable ones. The system cannot know you did not mean it.
It cannot see what is not written. A candidate who did not describe a capability does not have it as far as the comparison is concerned.
It reflects the shape of your posting. Narrow postings produce narrow candidate sets. If your matches all look the same and none of them are quite right, widen what you asked for before concluding the market is empty.
Frequently Asked Questions
Employer Learning Center
On GigFinder
From the candidate side
Post a role and see the matching
The clearer the posting, the more useful the ordering that comes back.