New Study Validates Cefr.app's Interactive AI Language Proficiency Test as Faster EFSET Alternative

HOUSTON - July 16, 2026 - PRLog -- New study demonstrates that an interactive, computer-adaptive AI language proficiency assessment achieves strong concurrent validity with established high-stakes exams while delivering results in a fraction of the time.

The study, titled "Comparing Concurrent Validity of Interactive AI Language Proficiency Tests with Legacy Examinations*, evaluated the cefr.app platform against the International English Language Testing System (IELTS) Academic and the Duolingo English Test (DET). Researchers analyzed data from 542 participants using a mixed-methods, within-subjects design.

Key findings included:
- Strong Pearson correlations between cefr.app scores and legacy measures: **r = 0.84** with IELTS overall bands and **r = 0.81** with DET overall scores (both p < .001).
- Superior ordinal classification accuracy: The adaptive AI platform showed a Quadratic Weighted Kappa (κw = 0.88) agreement with independent expert ratings — higher than traditional formats.
- Dramatic efficiency gains: cefr.app typically completes testing in 3–15 minutes, compared to ~2 hours 45 minutes for IELTS and 1 hour for DET, while maintaining or exceeding classification reliability against expert CEFR anchors.

The implications of these results extend beyond individual test-takers. By significantly reducing testing time and operational costs while preserving — and in some cases enhancing — alignment with expert human judgments, the adaptive AI approach offers a scalable solution for universities, language institutes, and professional certification programs facing growing demand for efficient proficiency assessment. The study also highlights the potential for improved accessibility, particularly for mobile-first users and those in resource-limited settings, without compromising psychometric standards.

"These results support the psychometric rigor of interactive AI-driven testing models," said lead researcher. "The adaptive platform delivers equivalent or superior alignment with human expert judgments in far less time, addressing key limitations of legacy systems such as high costs, logistical friction, and vulnerabilities to test-wiseness strategies."

Full Study Available
The complete paper is published on ResearchGate:
https://www.researchgate.net/publication/410169089_Compar...

This validation study was conducted under the auspices of the American Education Institute specializing in scalable digital testing solutions. AEI was formally known as the Center for Language Research out of UNM.

-----
About AEI
The American Education Institute advances practical, efficient, and psychometrically sound language assessment methodologies through innovative AI tools and rigorous validation research.

Head of Research & Development
Michael Brekke, Ed.D.
College of Education, Curriculum & Instruction
Texas Southern University, Houston, TX