For hiring managers

What is resume parsing? How applicant tracking systems turn resumes into fields, and where it goes wrong

On this page
  1. How does resume parsing work?
  2. What does a resume parser extract?
  3. Why do resumes fail to parse?
  4. How is resume parsing different from screening, matching and ranking?
  5. Does resume parsing raise legal or compliance questions?
  6. What should recruiters do about parsing errors?
  7. What should staffing agencies know about resume parsing?
  8. A copy-ready parsed-profile check
  9. Questions people ask

Resume parsing is the automated extraction of information from a resume file, such as the candidate's name, contact details, work history, education and skills, into structured fields in an applicant tracking system (ATS) or recruiting database, so the profile fills itself in and the candidate can be found by search. A parser reads text; it does not judge the candidate, though its output often feeds separate matching, ranking or screening features. As of November 2026, Greenhouse describes its parser as scanning "an imported resume" and auto-filling "appropriate fields," and lists resumes uploaded as images, graphics, tables, columned layouts and contact details in headers or footers among the reasons a parse fails; Lever's API documentation says image files "will not be successfully parsed."

How does resume parsing work?

  1. The file comes in. A candidate applies through the careers page, a recruiter uploads a resume, or resumes arrive by email or bulk import. Size limits apply: Greenhouse says it "can't parse resumes larger than 2.5MB" (Greenhouse, Unsuccessful resume parse, as of November 2026).
  2. Text is pulled out of the file. Word and text-based PDF files contain text the parser can read; a photo or scan of a resume is a picture. Lever's developer documentation says that where a file is meant to be parsed, "image files are not supported and will not be successfully parsed for information," although they are still uploaded (Lever API documentation, as of November 2026).
  3. The text is split into sections and fields. The parser decides which words are the name, which lines are a job, which dates belong to which job, and which part is education. Lever lists the fields it can fill from a resume as "name, headline, emails, phones, location, company, and links," adding that "no information is guaranteed to be extracted," and stores past positions with "company name, location, job title, job summary, and start and end date."
  4. The fields are written to the profile. Lever says that when a recruiter also types details in, "the manually provided information will always take precedence over the parsed information."
  5. A person checks it, or nobody does. When Greenhouse cannot parse a file, it shows that "the resume has only been attached to the candidate," and "you'll need to manually input the candidate's details into the fields."

What does a resume parser extract?

FieldExample (made up)What to check after the parse
Name and contact detailsDana Ruiz, [email protected], (555) 010-0199That the name is not a heading such as "Curriculum Vitae," and the email is the candidate's, not a reference's
LocationColumbus, OHWhether a metro area or "remote" was turned into the wrong city
Current company and titleNorthside Logistics, Operations SupervisorThat the most recent job, not the first one listed, became "current"
Past positions with datesShift Lead, 2021 to 2023Overlapping or missing dates the parser invented by attaching a date to the wrong job
EducationB.S., Supply Chain ManagementThat a certificate or course was not filed as a degree
SkillsForklift certified, WMS, ExcelSkills pulled from a job description the candidate pasted in, rather than from their experience
LinksPortfolio or professional profile URLDead or unrelated links

Where an ATS searches only the structured fields, a failed or garbled parse can make a strong candidate invisible in your own database; where it also indexes the resume text, a keyword search can still find them. Ask your vendor which your system does.

Why do resumes fail to parse?

Greenhouse's help center gives the most specific public list, as of November 2026. A parse can fail for:

  • "Resumes that include graphics, photos, or word art"
  • "Resumes that are uploaded as an image, rather than a document (such as a .docx or .pdf)"
  • "Complex resumes with tables, headers, and footers" and "Resumes that have a columned layout"
  • "Resumes with the name and contact information in the header, footer, or text box"
  • "A resume with spaces between the letters"
  • A "fake resume," one "that contains fake or poorly disguised data"
  • Files over 2.5 MB

The practical advice for candidates follows directly: send a text-based Word or PDF file, put the name and contact details in the body, and keep each job's title, employer and dates together in plain lines. For recruiters, the lesson is that a blank or garbled profile says something about the file, not about the candidate.

How is resume parsing different from screening, matching and ranking?

StepWhat it doesDoes it judge the candidate?
Resume parsingExtracts information into fieldsNo
Keyword or Boolean searchFinds profiles whose text contains the terms a recruiter typedOnly in the sense that the recruiter's terms decide who is found
Knockout questionsRejects or flags applicants by their answers to set questionsYes, by a fixed rule; see what is a knockout question
MatchingCompares parsed data with job criteria and labels the fitYes, as a recommendation
AI screening or rankingScores or orders applicants, sometimes with reasonsYes

Greenhouse shows the boundary in its own documentation. Its separate Talent Matching feature "extracts skills, job titles, years of experience, start/end dates of employment" from resumes and assigns each assessed candidate a "match strength" result; the company says "humans are required to make the decision for each candidate" and that the feature "does not automatically advance or reject candidates" (Greenhouse, Talent Matching Data Processing FAQ, last updated February 2, 2026). Parsing errors carry straight into that kind of step: a missed job or a wrong date changes the "years of experience" a matcher sees. The wider picture of AI in an ATS is on what is an applicant tracking system.

Not legal advice. This section summarizes the federal regulation, the New York City and Illinois texts and the agency FAQ linked below as published in November 2026.

  • Keep the original. Employers covered by Title VII, the ADA or GINA must keep "any personnel or employment record made or kept," including "application forms submitted by applicants," for one year from the record or the personnel action, whichever is later (29 CFR 1602.14). The parsed fields are not a substitute for the resume the candidate sent.
  • Watch the proxies in parsed fields. Graduation years can stand in for age and home addresses for race or national origin. Illinois makes it a civil rights violation, with respect to recruitment and hiring, "to use zip codes as a proxy for protected classes" when an employer uses AI (775 ILCS 5/2-102(L), in effect since January 1, 2026; see Illinois HB 3773). Do not build filters on those fields.
  • Parsing versus a covered tool. New York City's Local Law 144 excludes "a tool that does not automate, support, substantially assist or replace discretionary decision-making processes and that does not materially impact natural persons," including a "database, data set, or other compilation of data" (NYC Admin. Code 20-870). A feature that scores or ranks applicants for a specific job is a different question, worked through on what is an automated employment decision tool. The city's FAQ adds that the bias audit and notice rules do not apply when a tool is used to "scan a resume bank" of people who have not applied for a specific position (DCWP FAQ).
  • Hidden text gets parsed too. Parsers read the text in the file, including text a person cannot see. How candidates hide instructions for AI screeners, and how to check for it, is on resume prompt injection.
  • Offer another way to apply. An applicant who cannot use the upload or autofill flow because of a disability may need an accommodation in the application process; say how to ask for one on the careers page.

What should recruiters do about parsing errors?

  1. Open the original before acting on the profile. Compare the current title, dates and location with the file, especially before rejecting someone for "missing" experience.
  2. Fix errors in the ATS record. Corrections made in a spreadsheet or email are lost the next time someone searches; the field rules are on the ATS data hygiene checklist.
  3. Search the full text, not only fields. If your ATS indexes the resume text, a keyword search can find candidates whose fields came through wrong; Boolean patterns for that are on the role pages such as Boolean search strings for recruiters.
  4. Re-ask for a better file when it matters. A short, polite request for a Word or text-based PDF version is reasonable when the parse failed and the candidate is moving forward.
  5. Write your own notes separately. Parsed data is the candidate's claim; your screening notes are your evidence. Keep them in the notes or activity fields described in how to write candidate notes in your ATS.

What should staffing agencies know about resume parsing?

Agencies parse resumes from many sources and then send resumes on to clients, so a parsing error can travel further. Two habits help. First, keep the candidate's original file on the record and build the client version from it, not from the parsed fields, so a parsing error does not end up in a submittal; whether to strip contact details from the client copy is covered in should recruiters remove contact details from resumes. Second, when a client asks the agency to filter candidates on parsed data, apply the same rules you would to any screening criterion: job-related, the same for everyone, and never a stand-in for a protected trait.

A copy-ready parsed-profile check

PARSED PROFILE CHECK (example; run before search results, submittal or rejection)
Candidate: ______________  Req: ______  Source: ______________
Original file attached and opens?            [ ] yes  [ ] no -> request a new file
Name / email / phone match the file?          [ ] yes  [ ] fixed
Location correct (city, remote, relocation)?  [ ] yes  [ ] fixed
Current title and employer = most recent job? [ ] yes  [ ] fixed
Job dates attached to the right jobs?         [ ] yes  [ ] fixed
Education entries are real degrees/certs?     [ ] yes  [ ] fixed
Any filter used on graduation year or zip?    [ ] no   [ ] yes -> remove it
Hidden or white text found?                   [ ] no   [ ] yes -> see policy
Checked by: ________  Date: __/__/____

The check is an example to adapt to your ATS. Once a recruiter knows where their own system usually goes wrong, most of the boxes can be skipped for clean parses and checked only where a field looks off. For high-volume desks, see how to screen resumes faster.

Questions people ask

What is resume parsing in an ATS?

It is the feature that reads an uploaded resume and fills in the candidate's profile fields automatically, such as name, email, phone, location, current company, past positions, education and links. Greenhouse describes it as scanning an imported resume and auto-filling the appropriate fields; when a parse fails, the file is only attached and someone has to type the details in.

Why did my resume not parse correctly?

Usually because of the file or the layout. Greenhouse lists resumes uploaded as an image rather than a document, graphics or photos, tables, headers and footers, columned layouts, contact details placed in a header, footer or text box, spaces between letters, and files larger than 2.5 MB as causes of failed parses. Lever's API documentation says image files will not be successfully parsed.

Is resume parsing the same as AI resume screening?

No. Parsing only extracts information into fields. Screening, matching or ranking features then use those fields to compare candidates with a job, and that later step is where scoring, bias and AI-law questions arise. Greenhouse, for example, runs a separate Talent Matching feature that assigns a match strength, and it says humans must make the decision for each candidate.

Should recruiters trust parsed resume data?

Treat it as a draft. Vendors say no information is guaranteed to be extracted, and dates, titles and locations are easy to misread. Check the parsed fields against the original file before searching, submitting or rejecting, keep the original resume attached to the record, and correct errors in the ATS rather than in a side spreadsheet.

Does resume parsing count as an automated employment decision tool in New York City?

Extracting text into fields on its own is unlikely to, because the city's definition excludes tools that do not substantially assist or replace discretionary decisions, such as a database or other compilation of data. A feature that scores, ranks or filters applicants for a specific job is different and may be covered. Check each feature against the law's definition; this is not legal advice.