AI Resume Screening · Screening · Hiring Process · AI Recruiting

AI Resume Screening: How JD-Aligned, Seniority-Weighted Scoring Works

Keyword filters reject good candidates and pass weak ones. How JD-aligned, seniority-weighted resume screening produces a score with the evidence behind it, what to ask of any screening tool, and why a person should still make every call.

Aryan Vasishta· Co-founder, 100Networks··5 min read

AI resume screening reads each resume against the job description and scores whether the person's real experience meets each requirement, weighted by how senior the role is. Done well, it returns a score with the evidence behind it, and a person still makes the call. Traditional keyword filters do neither: they check whether particular words appear on the page. Here is how JD-aligned, seniority-weighted screening works, what to ask of any tool that claims to do it, and where it stops being the right tool.

Why keyword filtering fails

The classic filter is a string match. You tell it "must have Kubernetes", and it rejects every resume without the literal word, including the engineer who wrote "container orchestration at scale on GKE". It also passes the resume that listed "Kubernetes" once in a list of skills with no supporting experience.

That is two failures at once: false negatives (capable people filtered out) and false positives (keyword-matchers passed through). The filter runs before anyone looks, so nobody ever learns what was lost.

What JD-aligned screening does instead

JD-aligned screening starts from the actual job description, not a keyword list. It turns the role into a handful of criteria, decides how much each one matters, then reads the resume for evidence that each is met.

  • The role needs "scalability experience". The resume says "rebuilt the ingestion pipeline to handle 4M events a minute". That is strong evidence, even though the word "scalability" never appears.
  • The role needs "team leadership". The resume says "mentored three engineers and owned the on-call rotation". Evidence found.

The output is not a yes-or-no gate. It is a score with the evidence attached: which requirements are strongly met, which are thin, and where the gaps are.

Why seniority weighting matters

The same resume should not get the same score for every role. A candidate with two years of experience and real depth is a strong junior and a weak staff engineer, both at once.

Seniority weighting changes how much each signal counts:

  • For a senior role, scope, architectural ownership and leadership evidence weigh heavily.
  • For a junior role, trajectory and fundamentals matter more, and the absence of leadership is not a penalty.

So the score is calibrated to the role you are hiring for, not to a generic "good resume". This is the difference between screening and skimming; the broader argument is in skills-based hiring vs. resume screening.

Keyword filter vs. JD-aligned screening

Keyword filter JD-aligned screening
What it checks Do exact terms appear? Does real experience meet each requirement?
Handles synonyms and paraphrase No Yes
Calibrated to seniority No Yes
Output Pass or reject Score plus evidence per requirement
False negatives High Lower
Auditable Not really Yes, you can see why

The score is a starting point, not a verdict

The most important design choice is that the screen scores and a person decides. A good screen hands your team a ranked list with the evidence behind every score. It does not silently reject anyone. Resumes understate what people have done all the time, so even a missing must-have is better treated as a prompt to look closer than as an automatic no.

From there, the strongest applicants move to a deeper signal: a structured AI interview or a rubric-graded take-home.

What to ask any screening tool

If you are evaluating a screening product, these questions separate the real thing from a keyword filter with a new label:

  1. Where do the criteria come from? They should be built from your job description and editable by you, not a fixed template.
  2. Can you see the evidence? Every score should come with the lines of the resume that earned it, per criterion.
  3. Does it know the level? Ask how the same resume scores for a junior and a senior version of the role.
  4. Does it ever reject on its own? The right answer is no.
  5. What happens to hard limits? A missing must-have should be flagged and explained, not used to bin someone unseen.

Where 100Networks fits

For the roles a team fills itself, the 100Networks Workspace includes a Screener. It reads every applicant for a role against criteria built from the job description, weights each one by how much the role needs it, and shows a score with its reasons for each part of the role: what the person has, and what the role asks for. Anyone below the role's hard limits is flagged as a reason to look closer. The Screener ranks and explains; moving or rejecting someone is always your team's decision.

Screening has a limit, though: it only sorts the people who applied. For the hardest roles, the right person is usually employed and not applying anywhere. 100Networks reports that 72% of the people it places were not looking when they were found. Those roles are better handed to the 100Networks marketplace, where a specialist recruiter who has filled that role before works it with an in-house sourcing desk behind them, and you pay 8.33% of first-year salary only when the person joins. The difference between screening and headhunting is usually where a stuck role gets unstuck.

The bottom line

Keyword filters reward how a resume is written. JD-aligned, seniority-weighted screening asks whether the person can do the role, and hands your team a shortlist with the evidence to back every score. Run it on the roles that attract good applicants, keep a person on every decision, and send the roles that don't to people who go and find them.

Frequently asked questions

What is AI resume screening?
AI resume screening reads each resume against the specific job description and scores whether the person's actual experience meets each requirement, weighted by how much the role needs it and by the level the role is hired at. Done well, it returns a score with the evidence behind every part of it, rather than a pass or fail based on which keywords appear on the page.
How is AI resume screening different from keyword filtering?
A keyword filter checks whether exact terms appear on a resume. JD-aligned screening reads for meaning: it recognises that 'rebuilt the ingestion pipeline to handle 4M events a minute' is evidence of scalability experience even though the word never appears, and it gives less credit to a skill that is listed once with nothing behind it. The result is fewer good people filtered out and fewer keyword-stuffers passed through.
What does seniority-weighted screening mean?
It means the same resume is scored differently for a senior role than for a junior one. For a senior role, scope, ownership and leadership evidence count for more; for a junior role, fundamentals and trajectory count for more, and a missing leadership record is not a penalty. The score reflects the role being hired, not a generic idea of a good resume.
Should AI screening reject candidates automatically?
No. A screen should rank applicants and explain each score, and a person should make the advance-or-reject decision after reading the evidence. Even a missing must-have is better treated as a reason to look closer than as an automatic no, because resumes routinely understate what people have done.
Can resume screening find candidates who are not applying?
No. Screening only sorts the people who applied. For hard roles that is often the wrong pool: 100Networks reports that 72% of the people it places were not looking for a job when they were found. Reaching them takes outbound sourcing and a recruiter who can persuade them, which is a different activity from screening applications.
How does the 100Networks Workspace screen applicants?
In the 100Networks Workspace, the Screener reads every applicant for a role against criteria built from the job description, weights each criterion by how much the role needs it, and shows a score with its reasons per criterion. It flags anyone below the role's hard limits as a reason to look closer, never an automatic no, and it never rejects anyone on its own. Access is by request, and pricing is set with each team with no per-hire fee.