A shortlist you can understand.
WhichCVFits helps recruiters compare CV evidence against the same job criteria. It assists your review; it does not decide who gets hired.
1. Describe the role
Add the responsibilities, skills, experience, and qualifications that matter to the job. Optional must-have requirements are assessed separately from the weighted score.
2. Review the CV text
Your browser reads PDF, DOCX, TXT, and Markdown files locally. You can inspect and edit extracted text before evaluating. The queue supports 20 CVs, files up to 5 MiB, PDF documents up to 30 pages, and 50,000 characters per CV.
Scanned, unreadable, and password-protected PDFs cannot be evaluated from their images. Paste readable text instead. Original files are not uploaded by the normal workspace flow.
3. Choose your AI
Connect a verified Google Gemini, Anthropic Claude, or OpenAI API key. Choose an available supported model and, where supported, a reasoning effort. Your provider bills API use directly; account access and quota depend on that provider.
4. Understand the score
Four scores contribute to the overall match: hard skills (40%), relevant experience (30%), domain and impact (20%), and qualifications (10%). You can change these weights; positive totals are normalized to 100%.
A mandatory requirement needs affirmative CV evidence. If it is unmet or unclear, the result cannot be shortlisted and the score is capped at 45. Missing CV evidence is not proof that someone lacks a skill.
- 85–100: Top fit
- 70–84: Strong fit
- 50–69: Partial fit
- Below 50: Weak fit
Your chosen shortlist threshold applies only when all must-haves are evidenced. Eligible candidates appear before disqualified candidates, with scores sorted high to low.
5. Make the next conversation count
Read the supporting quotes, compare up to three candidates, and use tailored interview questions to investigate gaps. Quotes are checked against the submitted text, not independently fact-checked. AI judgments can vary between runs.
Provider failures are shown explicitly. A failed candidate receives no invented score; you can retry failed CVs without evaluating successful candidates again.