Resume Parsing FAQ

Resume parsing is the automated extraction of structured information from resume documents into a machine-readable format like JSON. The parser reads the raw text from a PDF, uses AI models to identify which sections correspond to contact details, work experience, education, skills, and languages, and then maps each piece of information to a standardized schema. Unlike template-based parsers that break when the layout changes, AI-powered parsing understands the semantic meaning of text, so it correctly identifies a job title whether it appears in a sidebar, a table, or a bullet list. The output is a clean JSON object you can store directly in your database or feed into an ATS.

Our parser achieves 99.2% field-level extraction accuracy across standard fields including first name, last name, email address, phone number, work experience titles, company names, and educational degrees. This figure is measured against a continuously updated test set of real-world resumes in various formats and languages. In practice, this means that for every 1,000 fields extracted, approximately 8 may need manual review. Standard contact fields (name, email, phone) parse at near-perfect rates, while more ambiguous fields like job title versus department name occasionally require context that the document does not provide. Highly artistic or infographic-style resumes may have slightly lower accuracy.

The API currently supports PDF files up to 1MB in size. PDF is the dominant resume format, accounting for roughly 85-90% of resumes submitted in online applications. The parser extracts text directly from the PDF structure, which means it works best with digitally-created PDFs (exported from Word, Google Docs, or LaTeX). Scanned image PDFs will have reduced accuracy because the text layer depends on the scanning software's OCR quality. If you receive resumes in DOCX or other formats, convert them to PDF before sending to the API. Most single-page PDF resumes are well under the 1MB limit.

The API returns a structured JSON object with the following fields: firstName, lastName, emailAddress, phoneNumber, location, linkedInLink, otherLinks (an array of portfolio or website URLs), workExperiences (an array of objects containing title, organization, duration, and description for each role), educations (an array with degree, school, and year for each entry), fluentLanguages (natively spoken languages), and languages (all languages mentioned including proficiency levels). Every response follows the same schema regardless of the input resume format, so your integration code can rely on a consistent structure.

A single-page resume is parsed in under 8 seconds on average. Two- to three-page resumes typically take 12-15 seconds, and longer documents (4-5 pages) take approximately 20-25 seconds. These times include PDF text extraction, AI-powered field identification, and JSON response generation. The API processes requests synchronously, meaning you receive the full parsed result in the same HTTP response. For most web applications, the sub-8-second response time for standard one-page resumes is fast enough for real-time user-facing workflows without needing a background job queue.

No, they are different steps in the hiring pipeline. Resume parsing extracts raw data from a document and structures it into fields like name, email, job titles, and skills. Resume screening takes that structured data and evaluates it against job requirements, scoring or ranking candidates. Parsing is a prerequisite for automated screening: you need structured data before you can apply matching logic. Our API focuses exclusively on the parsing step, giving you clean structured data that you can feed into any screening, ranking, or matching system you build or use.

The AI model is trained on resumes in multiple languages and correctly extracts fields from resumes written in English, Finnish, Swedish, German, French, Spanish, and other European languages. Field names in the JSON response are always in English regardless of the resume language (e.g., the field is always "firstName" even if the resume is in German). The languages and fluentLanguages fields capture whatever language skills the candidate lists. For best results with non-English resumes, ensure the PDF has a proper text layer and is not a scanned image without OCR.

Yes, the AI-powered parser handles most creative layouts including multi-column designs, sidebar layouts, tables, and colored backgrounds. It understands resume structure semantically rather than relying on positional rules, so it correctly maps a job title in a sidebar the same way it maps one in a traditional top-to-bottom layout. However, extremely artistic resumes (infographic-style with icons replacing text, or designs where text is embedded as images) will have reduced accuracy. Standard professional formats consistently yield the best results across all fields.