On this page
- What is resume parsing?
- What does a resume parser extract?
- How does resume parsing work?
- What types of resume parser are there?
- What makes Indian CVs harder to parse?
- How accurate is resume parsing, and how do you test it?
- What are the risks of relying on resume parsing?
- What should candidates know about resume parsers?
- Frequently asked questions
- A parser extracts; it doesn't judge. Matching and scoring happen afterwards, and both depend on the parse being right.
- Older parsers use rules and keywords; newer ones use machine learning and large language models that read layout and context.
- Indian CVs add their own challenges: CTC in lakhs, notice periods, DD/MM dates, and personal details you shouldn't screen on.
- Check accuracy field by field on 50 to 100 of your own CVs before you trust the output.
What is resume parsing?
Resume parsing (also called CV parsing or CV extraction) converts an unstructured document, usually a PDF or Word file, into a structured candidate record. Without it, an applicant tracking system holds a pile of files that can only be searched as text. With it, you can filter by years of experience, location or skill, and an AI resume screening tool can compare each candidate with the role. Parsing is the first step of screening; for the full sequence from CV to shortlist, see how AI resume screening works.
What does a resume parser extract?
| Field | Example | Common problems |
|---|---|---|
| Contact details | Name, phone, email, city | Details in headers or footers are skipped by some parsers |
| Work history | Employer, title, start and end dates | Two-column layouts mix up which dates belong to which job |
| Total experience | "6 years 4 months" | Overlapping jobs counted twice; gaps miscounted |
| Education | Degree, institution, year, marks | Abbreviations (B.Com, BBA, PGDM) not mapped consistently |
| Skills | SQL, Tally, inbound voice, KYC | Skills only mentioned inside job descriptions are missed |
| Salary and notice | "CTC 4.2 LPA", "notice: 30 days" | Lakhs, LPA and monthly figures confused; often missing altogether |
| Languages | English, Hindi, Tamil | Claimed proficiency isn't proof of spoken level |
How does resume parsing work?
- Read the file. Text is pulled out of the PDF or Word file. Scanned CVs and photos of CVs need optical character recognition (OCR) first, which adds errors.
- Understand the layout. The parser works out the sections (experience, education, skills) and the reading order. This is where columns, tables and text boxes cause trouble.
- Extract the fields. Names, dates, employers and skills are picked out of each section.
- Standardise. Dates become durations, job titles and skills are mapped to a common list, and locations are matched to known cities.
- Store. The structured record goes into the ATS or screening tool, ready for search and matching.
What types of resume parser are there?
| Type | How it works | Strengths | Weaknesses |
|---|---|---|---|
| Rule and keyword based | Patterns and dictionaries ("Experience" heading, date formats, a skills list) | Predictable, cheap | Breaks on unusual layouts and wording |
| Machine learning | Models trained on labelled CVs to recognise fields | Handles more variety | Only as good as its training data; weaker on formats it hasn't seen |
| Large language model | Reads the CV like a person and returns structured fields | Copes with messy layouts and context ("handled escalations" as a skill) | Can occasionally invent or misread details; needs checks |
What makes Indian CVs harder to parse?
- Salary formats. "4.2 LPA", "₹35,000 per month", "CTC 6 L" and "6,00,000" all need to land in one field, with the right unit.
- Notice period and joining. "Immediate joiner", "serving notice, LWD 15th Nov" and "90 days, negotiable" are common and matter a lot, but parsers rarely read them reliably.
- Dates. DD/MM/YYYY, "Apr'22 – Present" and financial-year references can flip months and days.
- Job titles. "CSE", "CCE", "tele-caller" and "process associate" can all mean a customer support role; "fresher" means no experience, not missing data.
- Personal details. Many Indian CVs list date of birth, marital status, religion, father's name or a photo. A parser may extract them; your screening should ignore them.
- Formats. Naukri-exported profiles, designed templates with two columns, and phone photos of printed CVs from walk-in drives all appear in the same pile.
Because salary, notice period and shift preference are so often missing or unclear, the best fix is not a better parser but a question. Hirin.ai asks these directly on WhatsApp or an AI voice call and writes the answers to the candidate's profile.
How accurate is resume parsing, and how do you test it?
There is no independent, current benchmark that compares parsers on Indian CVs, and vendors' own accuracy figures are measured on their own test sets. Test it yourself, field by field:
- Pick 50 to 100 real CVs from recent roles, including scans, two-column designs, Naukri exports and freshers.
- Write down the right answers for the fields you screen on: total experience, current employer and title, location, skills, education.
- Run them through the parser and mark each field correct, wrong or missing.
- Score accuracy per field and per format. A parser can be right on emails and wrong on experience, which is the field that matters.
- Decide what needs a check. Fields that fail often should be confirmed with the candidate or reviewed by a recruiter.
Parsing plus the questions CVs don't answer
Every CV read and scored with reasons, then notice period, CTC and location confirmed on WhatsApp or an AI voice call.
What are the risks of relying on resume parsing?
A parsing error becomes a screening error. If experience is miscounted or a skill is missed, the candidate drops down the list without anyone seeing why. Rigid filters on parsed fields make it worse: in the Harvard Business School and Accenture "Hidden Workers" study (2021), 88% of employers said qualified high-skilled candidates were screened out for not matching exact criteria. Keep hard filters few, show the parsed data to recruiters, and read a sample of rejected CVs each week. Under India's Digital Personal Data Protection Act, 2023, also limit what you keep to what hiring needs.
What should candidates know about resume parsers?
Simple layouts parse best: one column, standard headings (Experience, Education, Skills), dates on every job, and a text-based PDF or Word file rather than an image. Recruiters can share this guidance in job posts to improve the data they receive.
Frequently asked questions
What is the difference between resume parsing and resume screening?
Parsing extracts the facts from a CV into fields. Screening compares those facts with the job's requirements and ranks candidates. Screening depends on parsing being right.
Can resume parsers read scanned or photographed CVs?
Most can, using OCR, but accuracy drops with poor scans, handwriting and photos taken at an angle. Test these formats separately, especially if you run walk-in drives.
Is an LLM-based parser better than a traditional one?
It usually copes better with unusual layouts and wording, but it can occasionally misread or invent a detail. Whatever the type, check accuracy on your own CVs and show recruiters the parsed fields.
Does resume parsing work with my ATS?
Most ATSs include a parser or connect to one. Check which fields it fills, whether recruiters can correct them, and whether corrections flow back. See the recruitment software integrations checklist.
This guide describes how resume parsers work in general and the CV formats Indian HR and staffing teams see, based on Hirin.ai's AI screening as of October 2026. We found no independent, current accuracy benchmark for parsers on Indian CVs, so we give a testing method rather than a figure. Research was checked against its source on 12 October 2026.
- Harvard Business School and Accenture, "Hidden Workers: Untapped Talent" (2021), reported by HR Dive.
- Ministry of Electronics and IT, Digital Personal Data Protection Act, 2023.
- Hirin.ai, AI screening product page.
Vikas Agarwal
Vikas Agarwal is the founder of Hirin.ai, a powerful AI Recruitment Software powered by AI Agent - Zena, redefining how companies find and assess talent. With years of experience leading digital product innovation, he brings a sharp focus on solving real hiring challenges. Vikas likes to talk about AI, recruitment tech, and the future of work.