Stop fighting the search box. Start asking it questions.
Chosen gives you three ways to find candidates — natural language, boolean, and semantic — plus an AI match rating that doesn't just give you a number, but shows you the evidence behind it.
Your ATS search was designed for databases, not for recruiters.
You know who you're looking for. You could describe them in one sentence to a colleague. But your search box needs exact keywords, in the right fields, with the right filters, in the right order. And the match scores your ATS generates? A mystery number with no explanation. Is 72% good? Who knows.
Three ways to find candidates. All in one search box.
Smart search (natural language)
Type what you want like you'd say it out loud: "Senior backend engineers from NYC with 5+ years of Python who applied in the last 30 days." The AI breaks the sentence into semantic meaning and structured filters.
Boolean search
For recruiters who think in operators: (python OR java) AND NOT junior. Full AND/OR/NOT with parentheses. Parse tree returned so you can verify.
Semantic / vector search
Find candidates whose profiles are semantically related to a concept. "Engineers who've built real-time systems" finds matches even if none use those exact words.
Structured filters
Filter by stage, job, or any custom property with equals, contains, greater-than, less-than, is-empty. Save filter states as views.
Not a number. An evidence trail.
Claim extraction
The AI reads the job description and internal match criteria and extracts 5–15 discrete, testable claims — "3+ years of Python," "experience leading a team," "must have worked in regulated data environments." Each is classified as must-have, nice-to-have, or red flag with a weight from 1–10.
Evidence evaluation
The AI reads the candidate's resume and evaluates each claim individually, demanding concrete evidence. No hand-waving. No assumptions. Missing claims are treated as gaps, not definitive 'no's.
Transparent breakdown
You see every claim, its classification, its weight, the candidate's score on it, and the specific evidence from the resume that supports the evaluation. You see what they clearly meet, what they partially meet, and where the gaps are.
Defensible scores
Show the match breakdown to a hiring manager and point to exactly which requirements the candidate hits and which they miss. The score is never a black box.
Criteria, audit, and override.
Internal match criteria
Add a private match criteria block to any job. Encode unwritten rules — "ideally from a high-growth B2B SaaS," "non-negotiable: 5+ years with regulated data" — without exposing them on the career page.
Automatic recomputation
Edit the job description? Scores recompute for every linked candidate. Re-parse a resume? Scores refresh on every job. You never remember to re-run anything.
Manual override
The AI is a starting point, not the final word. Override any match score. Keep the override even as other data changes.
Audit trail
Every evaluation is stored and viewable. Perfect for compliance, DEI reviews, and interviewer calibration.
Stop coordinating. Start recruiting.
Book a 20-minute demo and see what happens when AI agents handle the logistics.