AI Writing Triage for HR & Recruitment Teams
EyeSift helps recruiting and talent acquisition teams screen written application materials—cover letters, take-home writing samples, and candidate questionnaires—for statistical AI-generation patterns locally in the browser, without uploading candidate PII to external servers.
Why Application Triage Requires Nuance
Generative AI tools are now standard in candidate preparation—from polishing resume bullet points to drafting full cover letters and take-home responses. Treating every AI signal as disqualifying risks rejecting strong candidates who used grammar assistants or resume templates, while ignoring AI signals entirely makes asynchronous writing assessments hard to trust.
How EyeSift Supports Recruiting Workflows
Zero-Upload Candidate Privacy
Resumes and cover letters contain names, phone numbers, addresses, and employment history. EyeSift runs text scoring entirely inside your browser tab so candidate PII never leaves your device.
Long-Form Cover Letter & Take-Home Screening
Evaluate 250–1,000 word cover letters, analytical memos, and written case studies using sentence-length burstiness, type-token diversity, and AI transition-phrase density.
Calibrated Caution on Bullet-Point Resumes
Resume action-verb fragments naturally exhibit low structural variance. EyeSift flags short or fragmented samples explicitly so recruiters do not mistake standard resume formatting for AI authorship.
Audio & Container Metadata Triage (Non-Biometric)
For asynchronous audio or video submissions, EyeSift inspects local container metadata, bitrate, C2PA markers, and basic waveform dynamics—never facial biometrics or voiceprint identity.
Free for All Hiring Managers & Panelists
No per-seat licensing fees, account walls, or procurement bottlenecks for occasional writing-sample checks.
Candidate Writing Assessment Rubric by Document Type
| Submission Type | Signal Reliability | Primary False-Positive Risk | Recommended Follow-Up |
|---|---|---|---|
| Resume Bullet Points | Low (~75–79%) | STAR-method templates, career-coach edits, and parallel action verbs | Ask for concrete project metrics, trade-offs, and technical architecture details in live screening |
| Cover Letters (250–500 words) | Moderate | Polished corporate register or ESL candidates using grammar-checking extensions | Check whether company-specific reasoning is genuine or generic boilerplate |
| Take-Home Writing Samples (500+ words) | Moderate–High | Formulaic executive-summary templates | Conduct a 10-minute live walkthrough where the candidate defends key assumptions and edits |
Defensible Governance & Compliance Best Practices
Never Use Heuristic Scores as an Automated Rejection Gate
Aligned with NIST AI Risk Management Framework (AI 100-1) and EEOC guidance on algorithmic employment selection, statistical writing scores must never automatically advance or reject a candidate without human review.
Publish Clear Candidate AI Guidelines Before Assessment
Specify in the job posting or take-home prompt whether grammar checkers, LLM outlining, or full drafting are permitted—especially for content, legal, engineering, and communications roles.
Account for Non-Native English Speaker (ELL) Variance
Stanford HAI research (Liang et al., 2023) demonstrated that perplexity-based detectors disproportionately flag non-native English writing. Always pair any statistical flag with a job-relevant live discussion.
Verify Domain Mastery Through Live Follow-Up Questions
The fastest way to validate a flagged resume or writing sample is to ask the candidate what failed during the project, what constraint forced a design change, and how they measured success.
Screen Written Applications Privately
Run browser-based heuristic checks with zero server retention and transparent signal breakdowns.