AI-assist detection

Detect AI use in interviews

Candidates increasingly read answers from AI tools like ChatGPT during remote interviews. Hiriso quietly flags when a spoken answer looks AI-generated โ€” giving you a neutral nudge to probe deeper with a follow-up. No proctoring, no surveillance, and never an automatic verdict.

How AI-use detection works

Compares answers to AI reference answers

For each question, Hiriso generates what a strong AI answer would look like from the same inputs โ€” then compares the candidate's spoken answer for matching substance and structure.

Raises a subtle, neutral flag

If an answer looks AI-generated, the Q&A item gets a quiet indicator โ€” no popups, no sound. It's framed as a prompt to ask a follow-up, never as an accusation.

Recruiter-only, click to reveal

The reference answer is hidden by default and visible only to you, so you can judge whether the candidate recited it. It never appears on any candidate-facing surface.

An overall AI-assist likelihood

The interview summary includes an aggregate AI-assist likelihood, with per-question detail โ€” conservative by default so it under-flags rather than over-flags.

Detection is a neutral assistant, not a judge. It helps you ask better questions โ€” it never decides for you. Pair it with Hiriso's unbiased scoring and full AI interview assistant.

Frequently asked questions

How does Hiriso detect AI use in interviews?

Hiriso generates a reference answer for each question and compares it to the candidate's spoken answer for matching substance and structure. When the similarity is high, it raises a subtle flag suggesting you ask a follow-up. It works on every interview regardless of how it's recorded.

Can it tell if a candidate is using ChatGPT?

It detects when an answer closely tracks what an AI would say โ€” which is what happens when someone reads from a tool like ChatGPT. It's a probabilistic signal to prompt a follow-up, not definitive proof, and it never auto-rejects a candidate.

Does it penalize well-prepared candidates?

The threshold is high and conservative by default, and timing is only ever a soft secondary input. Thinking time is legitimate. The goal is to prompt a human follow-up, not to punish strong preparation.

Is this proctoring or surveillance?

No. Hiriso does not monitor screens, verify identity, or record anything beyond the interview itself. Detection is a quiet, recruiter-only signal based on the answer's content.

Interview with confidence

See AI-assist detection in your own interviews. Free to start, no credit card.

Start Free