How to screen resumes without AI: a method that takes two minutes per candidate (2026)

A screening method for small teams that does not rely on resume scoring: decide the three requirements, knock out with questions, read for evidence not keywords, and record the decision. Faster and fairer than a model you cannot inspect.

Last updated 3 September 2026. Competitor pricing and plan limits are re-verified against each vendor’s own pages.

Screen resumes without AI by deciding the three hard requirements before opening a single application, checking them with yes/no or numeric screening questions, then reading each remaining resume for evidence of those requirements rather than for keywords.

Two minutes per resume is enough when you know what you are looking for; the time is lost when you do not.

Record the reason for every rejection in one line. That record is what makes the process consistent, and it is what an AI scorer cannot give you.

Step 1: write down the three requirements

Before the job is published, agree with the hiring manager on the three things a candidate must have, phrased so that a resume can show them: a licence, a number of years doing a named thing, a location. Everything else is a preference. Most screening takes too long because the screener is holding fifteen criteria and applying them inconsistently.

Step 2: ask the requirements as screening questions

Put the requirements that have a factual answer on the application form as yes/no or numeric questions, and mark the ones you would reject on as knockouts. Applicants who fail are filed into Screened out automatically and told at once. On ofper that is up to five questions per job, evaluated once at submission, with the answers kept. Nobody's resume is read by software; the applicant answered a question the employer wrote.

Step 3: read for evidence, not keywords

For each remaining resume, look for one piece of evidence per requirement: a role where they did the thing, for how long, with what result. A resume that lists the keyword without a role that demonstrates it is a no. A resume that demonstrates it under a different name is a yes. Keyword matching, whether by a person or a model, gets both of those wrong.

Requirement met, with evidence: shortlist
Requirement missing: reject, with the requirement as the reason
Unclear: shortlist for a screening call, and ask the question directly

Step 4: record the decision in one line

Move the candidate to Shortlisted or Rejected and write the reason: which requirement was met or missed. On a pipeline the move is dated and attributed automatically. A month later, when someone asks why a candidate was rejected, the answer exists. That is the whole audit trail, and it is what makes the process defensible.

Why not let AI score them?

Because you cannot see why it ranked a candidate where it did, cannot correct it when it is wrong about a particular person, and cannot show a candidate or a regulator the reason. A three-requirement checklist is slower per resume than a model by a few seconds and faster overall, because it needs no review of the model's mistakes. ofper has no AI screening, by design.

Questions, answered straight

How long should resume screening take?

About two minutes per resume once the requirements are written down. Forty applicants is an hour and a half, which is less than the time spent interviewing one wrong candidate.

What about applicants who do not meet the requirements but look promising?

Shortlist them for a screening call with a note saying why. The requirement list is a tool for consistency, not a rule against judgement; the note is what keeps judgement consistent.

Does ofper score or rank resumes?

No. ofper extracts resume text in the browser to pre-fill the candidate's details, applies the employer's own screening questions, and leaves the reading to a person.