Screening Best Practices

Metaview’s Screening lets you hear from every applicant through short AI-led interviews, with each completed interview assessed against rubrics you write — while keeping every decision with your team. It’s designed to widen who gets heard in your process, not to replace the people running it.

​

What Screening does — and doesn’t do

Screening asks your questions and applies your rubrics. It doesn’t decide.

• You write the interview plan: the questions, the optional follow-ups, and a rubric for each question describing what a strong answer looks like
• The AI interviewer conducts the conversation using your questions — it asks and listens, and doesn’t assess anyone during the interview
• Each completed interview’s transcript is checked against the rubrics you wrote, producing a rating with written reasoning and links to the exact moments in the conversation
• You review the scorecard, watch the recording where you want to verify something, and make every decision

Screening does not progress or reject candidates based on their interview. A rating on its own has no effect on a candidate’s application — candidates only move stages in your ATS when you click Progress or Reject yourself. The one automatic action available is Reject at deadline, which is off by default, only acts when you’ve explicitly turned it on, and is triggered by a missed deadline — never by a rating.

​

How it works

1. You build the interview. Draft the questions and rubrics for the role — Metaview produces a first draft from your job description and context, and you edit it until it reflects what you actually want to ask. Nothing goes to candidates until you publish.
2. Candidates interview on their own schedule. Enrolled candidates take the interview from their browser, with the AI interviewer asking your published questions.
3. Answers are checked against your rubrics. Each completed interview gets a scorecard: an overall rating (Great, Good, Okay, Poor, or Inconclusive), a summary, and per-question notes on how the answer matched your rubric — with links back to the transcript.
4. You review and decide. You decide whose interviews to look at more closely, who to progress, and who to reject.

If an interview comes back as Inconclusive, there wasn’t enough in the conversation to assess against your rubrics — usually a cut-short interview or a candidate who didn’t engage with the questions. It’s not a quality judgment, and you can offer another attempt with Reset attempt. For the full product walkthrough, see the Screening Overview.

​

Writing your questions and rubrics

The quality of the scoring depends directly on the quality of your rubrics. A vague rubric produces less consistent, less useful ratings.

• Describe what a strong answer contains. “Names a specific system they owned, the scale it ran at, and a decision they made about it” is more useful than “good technical depth.”
• Anchor to the job, not intuition. Ask about skills, experience, and situations directly relevant to the role, and write rubrics around role-relevant evidence.
• Write rubrics for the transcript — that’s all the rating sees. Ratings are produced from the transcribed conversation only: there is no video analysis, and no audio analysis beyond converting speech to text. Appearance, tone, and delivery can’t be scored, so a rubric that references them can’t be applied. Describe the substance a strong answer contains — which also treats candidates interviewing in a second language, with nerves, or with different communication styles the same.
• Avoid criteria that could introduce bias. Skip anything touching personal characteristics or signals unrelated to job performance. Screening has guardrails that block questions and rubrics referencing protected characteristics — but don’t lean on them: review your questions and rubrics with this in mind before publishing.
• Preview before you enroll. Preview as candidate is free and lets you take the interview exactly as candidates will. Then review your first few real scorecards against the recordings to check the ratings reflect what you’d conclude yourself — and refine your rubrics before enrolling at volume. When you edit and republish, scorecards are regenerated, so results stay comparable across candidates.

If a hiring team shares the screen, agree on the questions and rubrics together before publishing — and compare notes on borderline scorecards as you go.

​

Reviewing scorecards and making decisions

Every rating comes with reasoning and links to the exact moments in the transcript, so you can check how an answer was assessed against the interview itself — and watch the recording whenever you want the full picture. The recording is there for your review: the automated assessment reads the transcript and never watches or analyzes the video. Keep in mind that ratings reflect how the answers matched your rubrics, not the quality of the person. A strong candidate can rate lower simply because their answers didn’t surface evidence against your criteria — a reason to check the transcript before acting on a rating, and to keep refining rubrics that produce surprising results. Automation stays your choice, and it’s worth being deliberate about the two switches:

• Auto-enroll adds candidates from your ATS stage to the screen automatically. It automates the invitation, not any assessment or decision.
• Reject at deadline rejects candidates who don’t complete the interview in time. It’s off by default. If you turn it on, make sure your deadline is generous, your reminders have room to land, and your rejection email reads the way you’d want — this is the one place a candidate can be rejected without a person clicking the button.

​

AI use and misrepresentation signals

Screening checks each interview for signs that answers were read from an AI tool, and signs the person interviewed may not match the profile they applied with. Treat these exactly as the product presents them:

• They’re indicators, not conclusions. Each signal comes with reasoning and links to the relevant moments — review those moments before drawing any conclusion.
• They never change a candidate’s rating, and they should never be a rejection reason on their own.
• You can override a signal with your own assessment and justification after reviewing the interview — and you should, whenever what you see doesn’t support it.
• A flag is a prompt for a fair look, not a verdict. Well-prepared, coached, or non-native-speaker candidates can trip stylistic signals. Where something genuinely warrants it, the fairest next step is usually a conversation with the candidate.

​

The candidate experience

Screening interviews are candidate-facing, so how you configure them is part of how your company shows up.

• Be upfront that it’s an AI interview. The experience is designed so it’s obvious at every point that candidates are speaking with an AI, not a person — nobody is left to work it out for themselves. The invitation and welcome screen are candid about it, and the candidate guide explains the whole experience in plain language — link it in your own communications so candidates know what to expect and who sees their answers.
• Turn on the opt-out link in your invitation emails so candidates can decline AI screening, and have a path for what happens next — opted-out candidates shouldn’t simply disappear from your process.
• Think before requiring the camera. An active camera can be made mandatory, but consider accessibility and candidate comfort — audio and transcript alone are often enough to assess against your rubrics.
• Set humane deadlines. Candidates fit these interviews around jobs and lives; reminders are handled automatically, and a short heads-up email from your own address meaningfully improves completion.
• Be generous with second chances. Disconnections happen. Candidates can resume within a few minutes, and Reset attempt lets you offer another go when something went wrong.

​

Keeping things fair

Screening is designed to support fairer, more consistent screening than ad-hoc processes: every candidate gets the same questions, assessed against the same rubrics, with the same opportunity to make their case — rather than only the few who make it past a resume skim. It does not use protected characteristics in its assessment, and we run both internal and independent third-party bias testing on Screening, just as we do for Application Review. The same good practices you’d follow anywhere in hiring apply here too: ground your questions and rubrics in job-relevant skills and experience, apply consistent standards, review outcomes periodically to check things are tracking the way you’d expect, and make sure the team using Screening knows your company’s policies and any applicable local requirements.

​

Your responsibilities

A few practical steps sit with you as the employer, and requirements vary by jurisdiction — AI-led interviews are specifically regulated in some places. Screening was built with these regulations in mind. It’s designed to support compliance with rules like the EU AI Act, the Illinois AI Video Interview Act, and New York City’s AEDT requirements — and similar laws in Maryland, Colorado, and elsewhere. How you configure and use it still matters, though, so check the specifics with your legal team. Supporting documentation is in our Trust Center, and we’re always happy to help.

• Transparency. Many jurisdictions require that candidates are informed when AI is used in hiring, and some regulate AI interviews specifically. The invitation email, welcome screen, and candidate guide do much of this work; make sure your privacy policy and job postings are consistent with it.
• Offer an opt-out where required. Some jurisdictions give candidates the right to opt out of AI-assisted processing. The opt-out link is built into invitation emails — you need to enable it and make sure candidates know it’s available.
• Handle data subject requests. Candidate data originates from your ATS and from the interview itself, so if a candidate requests access to or deletion of their data, that obligation sits with you. Metaview provides the tools to support this.

​

Data and privacy

Screening uses the candidate data in your ATS and what the candidate provides in the interview itself — their recording, transcript, and answers. If the candidate’s camera is off, only the audio recording and transcript exist. Ratings are produced from the transcript of the conversation against your rubrics — recordings are stored for your playback, not analyzed — and no external data about the candidate is used. Your data is never used to train AI models, is not shared across customers, and is not used to generate or improve results for other customers. All personal data processing is conducted in accordance with our Data Processing Agreement, in compliance with GDPR and other international data privacy regulations with equivalent standards (including CCPA, PIPEDA, and LGPD). For more detail on data processing, AI governance, and security, please refer to our Trust Center. If you have any additional questions, you can reach us at privacy@metaview.ai.

​

You’re in control

Screening helps you hear from every candidate and puts a structured, evidence-linked scorecard behind each interview — but every decision remains yours. The AI asks your questions and applies your rubrics; you decide who to review, who to progress, and who to hire. If you have questions or need help getting started, our team is here to support you.