DOCS
Community open_in_new Support
  • Spanish (Mexico) check
  • Portuguese (Brazil) check
  • English (US) check
Login

Products

Case studies open_in_new

Acquire

  • Talent AcquisitionCreate requisitions and manage hiring
  • Candidate ExperienceDeliver personalized candidate journeys
  • AI InterviewerConduct and score AI interviews
  • Talent FlexMatch and contract gig talent

Develop

  • Talent ManagementGuide employee growth and mobility
  • Workforce ExchangeConnect job seekers with employers

Plan

  • Resource ManagementMatch employees to open projects
  • Reporting & AnalyticsBuild reports across talent data

Platform

  • Talent Intelligence PlatformPower talent decisions with AI
  • App platformREST APIs and webhooks
  • API docsEndpoint reference, authentication, and examples

What's new

  • Releases open_in_new
  • Articles open_in_new

The latest release is live

Granular audit logs, role-based export controls, and multiple SSO identity providers across the platform.

Read the release notes open_in_new
Products
  • New Home
  • Releases open_in_new
  • Articles open_in_new
  • Community open_in_new
  • Support
Language
  • Spanish (Mexico)
  • Portuguese (Brazil)
  • English (US) check
Login
  1. Home
  2. New Home
  3. Solutions
  4. Acquire
  5. AI Interviewer
  6. Understand Language proficiency evaluation
close /

No matching articles

Try a product name, or contact support.

homeAI Interviewer
  • Standalone AI Interviewer
    • AI Interviewer Overview
    • Languages supported by AI Interviewer
    • Candidate experience with AI Interviewer
    • Understand AI interview score
    • Schedule interviews using AI Interviewer
    • Understand AI Interview Guides
    • Create AI Interview Guides
    • Understand Language proficiency evaluation
    • AI coding interview scoring
    • Schedule an AI Interview
    • Work with Interview Library and Guides
    • Proctoring and flagged activity with AI Interviewer
    • 360 AI Interviews
    • Guide-level proctoring
    • AI Interview Insights in the pipeline
    • Automate AI interviews
    New Home

    Understand Language proficiency evaluation

    scheduleLast updated on November 13, 2026

    Overview

    Eightfold’s AI language proficiency evaluation provides you with automated, fair, and consistent insights into a candidate's speaking abilities during interviews. It seamlessly integrates into your interview process, using advanced AI to analyze a candidate's spoken responses and assign a clear language proficiency rating. It's designed to give you reliable, explainable, and bias-controlled evaluations for hiring at scale, ensuring your candidates meet the necessary language standards.

    What's CEFR?

    Eightfold has aligned its evaluation with the Common European Framework of Reference for Languages (CEFR) to evaluate language proficiency. This framework is a globally recognized, open, and universal standard for describing language ability.

    Why Eightfold uses CEFR

    Eightfold has selected the CEFR framework for several key reasons:

    • Globally accepted: CEFR is recognized worldwide, allowing for international comparability of scores.
    • Ideal for interviews: Unlike proprietary tests (which are best used for academic or professional certification), CEFR is specifically designed to assess natural, unscripted speech in real-life settings, making it perfect for job interviews.
    • Transparent and flexible: It provides a clear and transparent way to report language proficiency through descriptive "can-do" statements. You can freely adapt its levels to align with your business or job interview requirements.
    • No licensing fee: CEFR is in the public domain, meaning no licensing or purchase is required for its use.
    • Supports different languages: While currently focused on English, the framework can support various languages.

    How it works for you

    1. Easy setup: Recruiters will find an Evaluate English language proficiency option available during the AI interview setup.
    2. Where it applies:
      1. Currently, this feature is available for profile screening interviews only.
      2. It's only for English language evaluations.
      3. Candidates are informed both in their email invitation and at the start of the interview that their language proficiency will also be evaluated. Beyond this notification, the interview flow remains unchanged for them.

    Understand the evaluation results

    After an interview, you'll see a clear, comprehensive report displayed in the feedback form.

    • Overall rating: An easily digestible CEFR level displayed at the top, representing the candidate's general proficiency.
    • Detailed breakdown: We provide individual CEFR levels and explanatory feedback for four key dimensions of spoken language, giving you granular insights:
      • Pronunciation: How clear and intelligible their speech is, and how easily candidates are understood.
      • Fluency: How smoothly and naturally they speak, their pace, and how well they maintain conversation flow.
      • Grammar: Their accuracy and range of grammatical structures used.
      • Vocabulary: The breadth, precision, and appropriate use of words in their responses.

    Language proficiency calculation

    Our language proficiency evaluation system follows a structured process:

    1. Response collection: We collect audio responses from candidates during interviews, filtering out short clips and concatenating longer ones for coherent speech flow.
    2. Pre-processing: We normalize and improve the audio for analysis while removing noise artifacts.
    3. Analysis: We evaluate responses based on CEFR rubrics across four dimensions: Pronunciation, Fluency, Grammar, and Vocabulary. Each dimension is scored independently, providing detailed insights along with an overall CEFR proficiency level (A1-C2).
    4. Calibration: We apply a linear regression layer to improve exact score matches, increasing precision in our final output.

    Note: The evaluation is only performed on segments of a candidate's speech that are at least 15 seconds of uninterrupted speech.

    Common reference levels

    This table provides a summary of the common reference levels.

    Proficient user C2
    (Proficient)
    Can understand with ease everything heard or read. Can summarise information from different spoken and written sources, reconstructing arguments and accounts in a coherent presentation. Can express themselves spontaneously, fluently and precisely, differentiating finer shades of meaning in complex situations.
    C1
    (Advanced)
    Can understand and interpret complex texts and meanings. Expresses fluently with little hesitation. Uses language effectively in various contexts. Writes clear, structured text on complex topics with good use of connectors and cohesive devices.
    Independent user B2
    (Upper Intermediate)
    Understands main ideas in complex texts on various topics, including technical ones in their field. Interacts fluently with native speakers without strain. Produces clear, detailed writing on many subjects and discusses viewpoints on current issues, weighing pros and cons.
    B1
    (Intermediate)
    Can understand key points in clear, standard input about familiar topics like work, school, or leisure. Can write simple, connected text on familiar or personal topics. Can describe experiences, events, dreams, hopes, and ambitions. Can briefly explain opinions and plans.
    Basic user A2
    (Elementary)
    Can understand simple sentences and common expressions about everyday topics. Can talk about routine matters and share basic information. Can describe their background, surroundings, and immediate needs in simple terms.
    A1
    (Beginner)
    Can understand and use common words or simple phrases to meet specific needs. Can introduce themselves and can ask or answer basic questions about personal details. Can interact in a simple way provided the other person talks slowly and clearly.

    Accuracy of proficiency evaluation

    We've rigorously validated our system's accuracy using real-world datasets, demonstrating its reliability and precision:

    • High agreement: 96.8% of generated scores fell within a ±1 tolerance of the ground-truth scores (expert-assigned ratings).
    • No systemic bias: Distribution analysis confirmed a strong relationship with ground-truth and no significant systemic bias.

    This means the model provides consistent and explainable results aligned with international CEFR standards. Further, operational cost analysis indicates the process is cost-efficient and scalable, making it a viable solution for large-scale production deployment.

    #language assessment #cefr #english language evaluation #ai language eval

    Was this article helpful?

    thumb_upYes
    thumb_downNo
    rate_reviewGive feedback about this article

    On this page

      No sections

      Related

      AI Interviewer AI Interviewer Overview
      AI Interviewer Languages supported by AI Interviewer
      AI Interviewer Candidate experience with AI Interviewer
      AI Interviewer Understand AI interview score
      AI Interviewer Schedule interviews using AI Interviewer
      AI Interviewer Understand AI Interview Guides
      AI Interviewer Create AI Interview Guides
      AI Interviewer AI coding interview scoring
      AI Interviewer Schedule an AI Interview
      AI Interviewer Work with Interview Library and Guides

      Eightfold is the agentic talent intelligence company that gives every HR leader, recruiter, and people manager the intelligence to see, develop, and deploy the full depth of talent around them.

      Talent intelligence, human-led

      Eightfold Platforms

      • Talent Acquisition
      • Talent Management
      • Workforce Exchange
      • Resource Management
      • API Documentation
      • Release Notes

      About Eightfold

      • Blog
      • Careers
      • Contact Us
      • Events
      • Leadership
      • Partners
      • Press
      © Eightfold, 2026. All rights reserved worldwide. Privacy Policy Cookie Notice
      Follow Us :

      Definition by Author

      0
      0
      Expand