Resume Screening

How to screen resumes manually

Before opening the first resume, write down three lists: must-haves (a fail on any one is a decline), strong positives, and things to ignore. For a medical billing specialist, must-haves might be experience with insurance claims and familiarity with CPT and ICD-10 codes. Positives could be experience with denials management or a certification. Things to ignore: school prestige, formatting, unexplained gaps.

Then screen in one pass with a simple scoring sheet: must-haves met yes or no, a 1 to 3 rating on each positive, and a short note on anything you need to ask about in the phone screen. Experienced recruiters typically spend one to two minutes per resume for high-volume roles. More than that usually means the criteria are not clear enough.

Screen against the criteria, not against each other. Comparing resumes side by side pulls you toward whoever writes the most polished document, which is a weak predictor of performance in most frontline jobs.

How AI resume screening works

Resume screening software works in layers. First, resume parsing extracts structured data from the document: job titles, employers, dates, skills, education. Then the system compares that data to the job's requirements. Older systems match keywords. Newer ones use language models to recognize equivalent skills and titles (a cashier and a front-end associate are similar roles) and to rank candidates by predicted fit.

Used well, AI screening is fast and consistent, and it can surface candidates a tired recruiter would skip at resume 300. Used badly, it scales bad judgment. A model that learns from your past hiring decisions will copy whatever patterns are in them, including biased ones. This is the core concern behind AI bias in recruiting.

Practical guardrails: use explicit, job-related criteria rather than letting a model infer what good looks like from past hires; keep a human reviewer on rejections, or at least on a regular sample of them; and test outcomes for adverse impact by group.

Bias in resume screening

Research over many years has found that identical resumes can get different callback rates depending on the name at the top. Manual screening is exposed to affinity bias, halo effects from a known employer or school, and fatigue. Automated screening is exposed to biased training data and proxies, such as zip code, graduation year or gaps that correlate with caregiving.

Common defenses:

• [Blind hiring](/glossary/blind-hiring/) at the resume stage: hide names, photos, addresses and graduation years during first review.
• Structured scoring against written criteria.
• Skill-based alternatives: for high-volume roles, a short assessment can replace much of the resume review.
• Monitoring: compare pass rates by group at the resume stage at least quarterly.

Legal rules for AI resume screening

Several laws now apply specifically to automated screening tools:

• New York City Local Law 144 (enforced since July 2023): employers using an automated employment decision tool to screen candidates for jobs in the city must have an independent bias audit conducted within the past year, publish a summary of results, and notify candidates at least 10 business days before use.
• Illinois: amendments to the Illinois Human Rights Act effective January 1, 2026 prohibit employers from using AI in ways that have a discriminatory effect and require notice to employees and applicants.
• Colorado: the 2024 Colorado AI Act was repealed and replaced in May 2026 by SB 26-189, which from January 1, 2027 requires notice, an explanation within 30 days of an adverse decision, data correction and human review when automated tools materially influence employment decisions.
• EU AI Act: classifies AI systems used to screen, filter or rank job applicants as high-risk, with obligations for providers and deployers.

Federal anti-discrimination law applies regardless of whether a human or software made the decision. Rules vary by state and country and are changing quickly.

Resume screening vs candidate screening

Resume screening is one method inside the broader process of candidate screening, which also includes knockout questions, assessments and phone screens. In professional hiring, the resume often carries most of the screening weight. In high-volume hourly hiring, it should carry less: many strong frontline candidates have thin or no resumes, and availability, location and a short skills check predict more than a polished work history.

Common mistakes

• Rejecting for gaps, short tenures or job hopping without asking why.
• Exact keyword matching that rejects equivalent experience described in different words.
• Requiring a resume at all for hourly roles where it adds little; it lowers application completion.
• Letting AI tools reject candidates with no human review and no audit trail.
• Inconsistent criteria between recruiters working the same requisition.