Recluto
Back home

AI disclosure

Recluto uses AI at multiple steps of the hiring process. We think you should know exactly where, why, and what your rights are. Here's the plain-English version.

Last updated · May 1, 2026

Many companies that hire with Recluto operate in jurisdictions with specific AI-decision laws — including the EU AI Act (high-risk hiring systems, in force from August 2026), New York City Local Law 144 (Automated Employment Decision Tools, audited annually), Illinois' AIVI Act, and Colorado SB 24-205. The disclosures below are designed to satisfy the candidate-information requirements under each of those laws and to give you genuine clarity about what's happening behind the scenes.

Where we use AI

CV screening

An LLM parses your uploaded CV into structured fields (employment history, education, skills, languages) and compares it to the job's requirements. The output is a 0–100 fit score and a recommendation (HIRE / MAYBE / NO_HIRE). Used as one input among many — never the sole basis for rejection.

Written assessment

If the role uses a written stage, an LLM may generate questions tailored to your CV, and may score your answers against rubric criteria the company configured. You can request to skip and go straight to a human reviewer at any point.

Voice interview

If the role uses a voice stage, you'll have a live phone or VoIP conversation with an AI interviewer. The conversation is recorded, transcribed, and scored. You can ask the AI to escalate to a human at any time, or hang up and reply to the email to request a human round.

Live video interview

If the role uses a video stage, you'll join a live video room with an AI interviewer that responds in real time. The recording and transcript are stored on the company's account. Same escalation rights as voice.

Final review

All AI outputs are visible to a human recruiter at the company you applied to. The hiring decision rests with the recruiter, never with Recluto.

What we don't do

  • We do not auto-reject candidates based solely on the AI score by default. The default in every workspace is recruiter review of every recommendation.
  • We do not use facial-expression analysis, gaze tracking, voice-tonality biometrics, or any "emotion AI" in our scoring. Scoring is based on the substantive content of your answers — what you said, not how you said it.
  • We do not use protected-characteristic predictions. The AI is instructed not to infer or use gender, age, ethnicity, religion, disability, sexual orientation, or pregnancy status, and our prompt and model evaluations are audited for this.
  • We do not share your data with ad networks, sell it to third parties, or use it to train foundation models without the company's explicit per-workspace opt-in (default is opt-out).

Your rights

  • Opt out of AI evaluation: at the start of every screening flow you can request a human-only review. The company you applied to receives the same shortlisted material with the AI score replaced by an "opted out" flag.
  • Escalate to a human at any time: during the written, voice, or video stage you can ask "I'd like to speak to a human" or click the escalate button — the company gets notified and a recruiter takes over.
  • Request the reasons: if you receive a rejection notice, you can reply asking for the basis. Under GDPR Art. 22 and similar laws, you have the right to a human-readable explanation of any automated decision, plus a route to contest it.
  • Request your data: ask for a copy of everything Recluto holds about you, or ask us to delete it. See privacy policy for the process — we respond within 30 days (typically within 5).
  • Skip recording: you can request that voice or video sessions not be recorded. The session may still be scored from a live transcript, but no audio/video archive will be stored.

Exercise any of these by replying to the email the company sent you, or by writing to privacy@recluto.ai with the company name and the email you applied with.

Foundation models we use

Our AI features run on a small number of state-of-the-art foundation models, accessed via OpenRouter (text reasoning + scoring), VAPI (voice telephony), Deepgram (speech-to-text), xAI (live video LLM), and Cartesia (text-to-speech). The specific model used at each stage is chosen for accuracy and latency and may change over time as models improve. We do not host our own foundation models.

How we monitor for bias

  • We instruct every model not to reason about or use protected characteristics in scoring.
  • We sample random batches of decisions monthly and run them through fairness checks: same-CV-different-name tests (Latanya / Lakisha / etc.), age-marker swaps, and accent variation tests on the voice stage.
  • Companies hiring under NYC Local Law 144 receive an annual independent bias audit summary, which we make available in Settings → Compliance for their workspace.
  • We document model upgrades in the changelog. Material changes to scoring logic are flagged.

If something feels wrong

We take complaints about the AI seriously. If a Recluto-powered interaction felt unfair, confusing, or biased — even if you weren't ultimately rejected — please write to privacy@recluto.ai. We investigate every report and respond with what we found, including changes we make to prompts, prompts we ship for the next model upgrade, or sub-processors we're reconsidering.

For comprehensive technical detail, see our security page and Data Processing Agreement.