Sapia.ai has released the third version of its AI-generated content detector, which the company says identifies nearly 97% of AI-written answers from the latest large language models.
The company says fewer than 1% of interviews on its platform are flagged as written entirely by AI, arguing that a trusted candidate experience can reduce the incentive to rely on generated responses.
Sapia.ai’s structured, text-based interviews are untimed and mostly completed by phone. If three or more answers are flagged as AI-written, candidates receive a real-time nudge rather than an automatic rejection.
“When candidates trust the experience, almost all of them answer in their own words.”
Barb Hyman, CEO and Founder, Sapia.ai
The latest model was trained using answers from current large language models and genuine candidate answers written before ChatGPT was released. The company says it tests and retrains the detector as new models are released.
Sapia.ai said it does not use webcam monitoring, eye-tracking or lockdown browsers, and that later human-led stages of recruitment should test what candidates describe in their initial interviews.
