How does Intrvio score language fluency and answer authenticity?
Intrvio's AI fluency scoring rates a candidate's spoken language on a CEFR-aligned scale from the live interview, and separate answer authenticity signals flag answers which may be LLM-assisted, memorized, or read from a script. Both are evidence for a human reviewer, not an automated pass or fail.
Fluency measured from real conversation
Language fluency is a real job requirement for many roles, and a self-reported level on a CV rarely matches reality. GAIA assesses spoken fluency from the live interview and maps it to a CEFR-style scale, so a recruiter sees a grounded read rather than a candidate's own estimate.
Because the score comes from a two-way conversation with follow-ups, it reflects how a candidate handles unscripted questions, not just a rehearsed introduction.
Authenticity signals for the LLM era
Candidates can now feed an interview question to a language model and read back a polished answer, which makes pure content quality a weaker signal than it used to be. Intrvio surfaces authenticity signals, such as answers that pattern-match generated text, unnatural pacing, or a mismatch between fluency and answer sophistication.
These signals do not accuse anyone. They flag answers that may need a closer look, so a reviewer can probe further or weigh the result with context rather than taking a too-perfect answer at face value.
Evidence, not an automated verdict
Fluency ratings and authenticity flags are presented as evidence alongside the transcript, never as an automatic decision. This matters because language and authenticity scoring touch fairness directly, and a human must own any conclusion.
Keeping a person in the loop is also what EU AI Act deployer duties and KVKK and GDPR expectations require, so the design treats these scores as decision support, not gatekeeping.
Fair by design
Fluency scoring is tuned to assess communication, not accent or dialect, and authenticity signals are calibrated to avoid penalizing strong but genuine answers. The aim is to reduce both false confidence and false suspicion.
When a signal is ambiguous, the design favors routing the candidate to human review over any silent downgrade, which keeps the process explainable to candidates and auditors.