Resume Parsing: How It Works and How to Check for Errors

Resume parsing is how an applicant tracking system (ATS) turns an uploaded resume into a searchable candidate record. Instead of typing details in by hand, you can use the extracted job titles, employers, dates, education, and skills to find and organize candidates.

The information doesn't always transfer correctly, though. A parser might miss a skill, attach dates to the wrong employer, or leave a job title blank. Those errors can change which candidates appear in your searches and what hiring managers or clients see in a profile. It helps to know where your parser goes wrong before a search depends on it.

What is resume parsing and how does it work?

Resume parsing is the automated process of pulling information out of a resume and organizing it into structured fields in a recruiting database. Parsers differ in how they work, but most combine some version of these steps:

• Extract the text: The software reads the text from a Word document or digital PDF. A scanned document may need optical character recognition (OCR), which converts an image of the page into readable characters.
• Identify sections: The parser uses headings and formatting clues to separate work experience, education, skills, and contact details.
• Pull out details: It picks out employer names, job titles, employment dates, certifications, and qualifications.
• Organize the data: The details fill fields in the candidate record. Some systems also standardize titles and skills so different terms land in the same search.

A parser might recognize "RN" and "registered nurse" as related terms, for example. Whether those terms return the same search results depends on the ATS and its search settings.

Resumes don't follow a standard format, which is where most errors start. Researchers in an October 2025 arXiv paper named the wide variety of resume layouts and content as a major challenge for automated extraction. A plain single-column resume is often easier for a parser to read than one with text boxes, complex tables, or multiple columns. A scanned file adds the chance of OCR mistakes.

Once extracted, the details become part of the candidate record in your applicant tracking system. On agency desks, they may also feed a recruitment CRM, where you search, filter, and manage candidates.

Which parsed fields should recruiters check?

Parsing errors don't all carry the same weight. A missing phone number can stop you from reaching a candidate, while a wrong job title can keep someone out of a relevant search. Start with the fields that affect finding, contacting, and qualifying people.

Pay particular attention to titles and skills if your system standardizes them. A warehouse employee whose title was "team lead" might be categorized as a supervisor, even if the role didn't involve managing people. That matters when you're searching for supervisory experience, so compare the standardized title with the resume before you rely on it.

How to test a resume parser before you rely on it

The most reliable way to judge a resume parser is to run the kinds of resumes your team receives through it. Consider an agency recruiter preparing to move an existing candidate database into a new ATS. The team places nurses, warehouse supervisors, and software engineers, so the test needs to show how the system handles resumes from each specialty.

Start with a small sample that covers different layouts and file types:

• Two-column resume: Check whether titles, employers, and dates stay together when they sit in separate columns.
• Scanned PDF: Look for OCR errors in names, contact details, and dates.
• Long federal-style resume: Check whether the parser separates many positions with lengthy descriptions.
• Resume with an employment gap: Confirm that dates stay with the right roles and no period is stretched to cover the gap.
• Non-U.S. CV: See how the parser handles unfamiliar layouts, day-first dates, and international phone numbers.

Open each parsed record beside the original document and compare the fields from the table above. Count something as an error when a field is missing, wrong, or attached to the wrong role. Differences in formatting, such as abbreviations or capitalization, matter less unless they change search results.

A simple tracking sheet is enough to see patterns:

One missing phone number may be a one-off. But if every scanned PDF loses its contact details, plan a manual pass on that part of the backlog before you import it. Give the most attention to errors that could change who appears in a search or whether someone looks qualified.

If you're evaluating a new vendor, ask to parse resumes you chose in a demo or test environment. A demonstration built on the vendor's own sample resumes won't show how the software handles yours.

How to find candidates whose resumes were parsed incorrectly

A candidate can have the right experience and still be hard to find because the parser didn't capture their resume. Suppose you're sourcing registered nurses with telemetry experience. A qualified candidate's resume mentions both, but the parser missed the skills section, so a search on structured skill fields alone leaves that person out.

If your ATS can search the original resume text, run a keyword or Boolean search alongside your usual field filters. For the nursing search, that could be telemetry AND (RN OR "registered nurse") across the resume text. Not every system offers full-text search, so check what yours supports.

This matters most in an older database, where records may have come from another system or been parsed with earlier settings. The same approach helps when you run a broader resume search across those records. Full-text search has limits too: if text extraction failed entirely, the missing words won't show up in results, and you'll need the original file.

Check the original resume before ruling out a candidate

A sparse record doesn't always mean the person lacks relevant experience. If a promising candidate has a blank skills section or an incomplete work history, open the original resume before you pass. The information may be in the document even though it never reached the fields.

An unexplained date discrepancy isn't evidence of a fake resume either. Check whether the parser attached the dates to the wrong position before you raise it with the candidate. When you find an error, correct the fields your team uses for searching and submissions, starting with active candidates and records you're about to share.

Resume parsing vs. resume screening

Resume parsing extracts and organizes information, while resume screening evaluates candidates against job requirements. A parser might record that a candidate has several years of accounting experience and lists a bookkeeping system as a skill. A screening tool might then use those fields to decide whether the candidate meets the requirements for an accounting role. Some systems go further by scoring or ranking candidates.

Screening tools are common. SHRM's 2025 release reported that 43% of organizations use AI in HR functions. It added that 51% use it "for recruitment tasks such as resume screening and candidate communication" without saying which group that share describes. Those findings cover AI in recruiting broadly, not parsing specifically.

Knowing which step failed helps when a candidate goes missing. If someone doesn't appear in your results, first check whether their details were extracted correctly. If the details are there but the person was filtered off a shortlist, the cause is in screening. Review the criteria, any automated filters or rankings, and your screening software settings with the hiring manager or client.

When automated screening raises legal questions

Scoring and ranking tools can bring legal obligations that basic extraction doesn't. New York City's Local Law 144 bars employers and employment agencies from using an automated employment decision tool without a bias audit within one year of its use. The audit information must be public, and candidates must receive certain notices.

Not every ATS feature is covered. Whether a screening or ranking feature counts depends on how it works and how it's used, so confirm the requirements with counsel before you rely on automated scoring. If you're weighing that kind of tool, the guide to AI resume screening covers what to check.

Why you should keep the original resume

The parsed record is useful for searching and managing candidates, but it isn't a substitute for the document the candidate sent. Keeping both lets you verify details when the fields look incomplete, and it makes errors easier to trace after a migration. Before you move records into a new system, confirm that the original files transfer with them, then open a few imported profiles to check.

Retention rules also apply to hiring records. The U.S. Equal Employment Opportunity Commission (EEOC) recordkeeping summary says private employers must keep hiring records, including applications, for one year. The clock runs from the date of the record or the personnel action, whichever is later.

Longer periods apply once a discrimination charge is filed, and other rules may apply by employer and location. Agencies should confirm which obligations cover their own records instead of assuming one period fits every client.

Start with five resumes before you trust the fields

You don't need to review every field in every old record to make parsed data dependable. Learn where your parser makes mistakes first.

Pull five resumes from roles you fill often and compare each parsed record with the original. Then note whether repeat errors come from the file format, the parser, or the field mapping. That tells you which records need a closer look and where a full-text search or a quick check of the original keeps a qualified candidate in view.