“Why this candidate?” has an answer.
Instead of pointing at a fit score, you point at the sentence in the resume that backs each must-have. The conversation moves from arguing with the tool to evaluating the candidate.
No more “trust the AI”. Every AI verdict in ShortlistTable links to the exact resume sentence behind it. Click the cell to jump to the highlighted PDF, override the verdict with one click, ship a shortlist with the receipts attached. AI screening, with the audit trail recruiters and legal teams actually need.
“Built and operated Kafka streaming pipelines at Ledger from 2022-present. Owned primary on-call for the payments event bus.”
The single biggest reason recruiters distrust AI screening is the lack of a clickable evidence trail. A model says “strong match” — based on what? Which sentence? Which page? If you cannot point at it, the verdict is hearsay.
Source-backed screening solves this by making the citation a first-class part of every cell. The verdict and the source sentence ship together, in the UI, in the audit log, and in the export. Disagreement becomes a productive conversation about the resume, not a stalemate about the algorithm.
Instead of pointing at a fit score, you point at the sentence in the resume that backs each must-have. The conversation moves from arguing with the tool to evaluating the candidate.
If a candidate or regulator asks why someone was screened out, you can produce the criterion, the verdict, the source sentence, and any subsequent recruiter override — with timestamps.
When a verdict feels off, you click into the cell, see which sentence the engine matched against, and adjust either the column wording or the cell value. No mysterious model behaviour.
Not “based on the resume” — the actual sentence. Lifted from the source document, attributable to a page and section.
Click any cell, jump to the source resume with the sentence highlighted in place. No copying, no searching, no doubt.
Every CSV / XLSX preset includes evidence columns next to verdict columns. The hiring manager sees both, your audit log sees both.
If the resume is silent on a criterion, we mark the cell as needs-review rather than fabricating a verdict. No citation, no answer.
When the recruiter changes a verdict, the original AI value and the source sentence both stay in the audit history alongside the override.
If you add a column or change the must-haves, we re-run only the affected cells against the same source documents. Citations stay stable.
| Property | Black-box AI | ATS resume scoring | Spreadsheet | ShortlistTable |
|---|---|---|---|---|
| Cell-level citations | ✕Rare | ✕None | –Manual notes | ✓Every cell |
| Click-to-PDF highlight | ✕No | ✕No | ✕N/A | ✓Sentence highlighted in source |
| Evidence in exports | ✕Score only | ✕Rank only | –Manual notes | ✓Evidence column per criterion |
| Override preserves history | ✕No | –Limited | ✕Last write wins | ✓Full timestamped trail |
| Verdict when resume is silent | ✕Often guesses | ✕Often guesses | ✓Blank | ✓Needs review |
The cell is marked “needs review” with the reason. We don’t fabricate a citation — if the resume is silent, we say so explicitly and surface the cell to the reviewer.
Page, section, and the exact sentence. For PDFs, we also store the bounding box so the highlight stays accurate when you click through.
Yes. Every screening column gets a paired evidence column in the export. A hiring manager opening the CSV sees the verdict alongside the source sentence.
Two paths: edit the column wording to be more specific, or override the cell directly with a recruiter note. Both edits are logged.
PDF, DOCX, TXT, and pasted text. PDFs have the richest citation (page, section, highlight); DOCX and TXT cite the matched line.
Yes — the candidate report preset bundles each verdict with its quote, source, and any recruiter override into a single PDF.
Every verdict carries the sentence behind it. Try it on 25 resumes — no credit card.