Fewer than 1% of interviews on Sapia.ai’s platform are flagged as written entirely by AI. The figure comes as Sapia.ai releases v3 of its AI-generated content (AGC) detector, which identifies nearly 97% of AI-written answers from the latest large language models (LLMs).
Sapia.ai’s structured, text-based interview is untimed and mostly completed on a phone, and every candidate gets personalised feedback at the end. Across over 10 million candidates, it averages 9.0/10 for satisfaction, and 86% choose to leave a comment about the experience.
“Everyone is talking about candidate fraud, and the reflex is to watch people harder, at the expense of the candidate experience.” said Barb Hyman, CEO and Founder of Sapia.ai. “Our data points elsewhere. When candidates trust the experience, almost all of them answer in their own words. That matters for the employer, because the interview is the first real experience a candidate has of your brand. The AI detector is there to nudge, never punish. Nobody is rejected automatically because of it.”
Sapia.ai scores the competencies candidates demonstrate in the interview, through the behaviours and experiences they describe, not how fluently they write. Using AI to polish an answer gives no real advantage. What matters is that the story is the candidate’s own. AI never makes the final hiring decision: later stages of the process, led by people, can test what candidates describe in the initial interview.
What’s new in v3
The latest version was trained on answers from a number of current LLMs from providers such as OpenAI, Google, Anthropic, xAI, DeepSeek, Moonshot AI and Alibaba, each paired with genuine candidate answers written before ChatGPT was released. Tested side by side on the same answers, v3 detects nearly 97% of AI-written answers from the latest LLMs, up from 92% for the previous version of the detector. Because new large language models arrive every few months, Sapia.ai tests the detector against new releases as they appear and retrains it as the landscape shifts.
How it works
Detection reads the answer itself. There are no intrusive methods like webcam monitoring, eye-tracking, or lockdown browsers. Before the interview starts, every candidate sees a note on AI use and commits to answering in their own words. If three or more answers in an interview are flagged as AI-written, the candidate gets a real-time nudge. A genuine answer is wrongly flagged around 2% of the time, but in a five-question interview the chance of an honest candidate having enough answers wrongly flagged to trigger a nudge is roughly 1 in 13,000. False-flag rates are similar across genders and ethnic groups.
About Sapia.ai
Sapia.ai is an AI interview platform that helps employers hire fairly and at scale through a structured, chat-based interview. Founded in 2018 by former CHRO Barb Hyman, Sapia.ai works with leading employers including Tesco, Nestlé and Woolworths Group across the UK, Australia, the Middle East and North America.
View source version on businesswire.com: https://www.businesswire.com/news/home/20261109164428/en/
Media gallery
