Tools to detect AI use in interviews: what they catch and what they miss

As of October 2026, tools to detect AI use in interviews fall into three groups. Technical assessment platforms (HackerRank, CodeSignal, CoderPad) flag copy-paste, tab switching and patterns linked to AI help inside their own tests and coding interviews. Live-call tools work in ordinary video interviews: Sherlock joins Zoom, Meet and Teams to flag hidden AI assistance, deepfakes and proxy candidates, and Pindrop Pulse for Meetings, Reality Defender and GetReal focus on deepfake voices and video. AI interviewer platforms such as HeyMilo build proctoring into their own screens. None of them proves a candidate cheated; a flag is a reason to ask a better question.

We make Interview Signal, one of the tools below; everything about the other products comes from their own sites, linked and dated. For identity fraud more broadly, including fake applicants and proxy interviewers, see our guide to spotting candidate identity fraud.

What kinds of AI use can these tools detect?

• Hidden AI help in a live interview: an app listening to the call and showing the candidate answers on screen or a second device.
• Read-out answers: fluent, generic answers that sound written rather than remembered.
• AI-generated code or test answers: pasted solutions, unusual typing patterns, similarity to known answers.
• Proxy candidates: one person interviews and another turns up for the job, or someone else answers off camera.
• Deepfakes: synthetic video or a cloned voice standing in for a real person.

Before buying, decide what you actually prohibit. Many employers now allow AI in some assessments and want to see how candidates use it, which is why HackerRank's homepage says integrity "isn't about whether you use AI or not" but about following the rules (HackerRank).

How do the detection tools compare?

As published by each vendor, as of October 2026. "Not published" means we could not find it on the vendor's own site.

Sources, checked October 2026: HackerRank plagiarism detection and Interview; CodeSignal cheating and fraud and pricing; CoderPad; Sherlock; Pindrop Pulse for Meetings and fake candidates; Reality Defender hiring and homepage; GetReal staffing and recruiting; HeyMilo; our pricing. "Not checked" means we did not review that vendor's integrations for this page.

Which tools work for technical assessments and coding interviews?

HackerRank

HackerRank's plagiarism detection uses copy/paste tracking, tab proctoring and image analysis to spot someone taking a test "with a buddy," and says it shows signs of using tools like ChatGPT; it claims a 93% accurate plagiarism model (HackerRank), a vendor figure without published method. In live interviews, integrity signals notify interviewers about clusters of copy-paste, window resizing and tab switching, and a Screen-to-Interview Identity Match uses facial recognition to compare the test-taker with the interviewee (HackerRank Interview). Best for: large technical hiring programmes. Watch: facial recognition raises biometric consent questions (below).

CodeSignal

CodeSignal's Suspicion Score combines solution similarity, telemetry and copy-paste activity into one trust score, its AI proctoring records video, audio and screen, and it verifies a government photo ID at the start of an assessment or interview. It says "human reviewers" validate flagged activity and final decisions (CodeSignal). Build is $79 a month billed annually and lists AI proctoring and ID verification (pricing). Best for: teams that want flags checked by a person before anyone sees them.

CoderPad

CoderPad offers real-time alerts on suspicious behavior, playback reports to confirm them afterwards, and, where AI use is allowed, a view of the candidate's prompt history and code edits (CoderPad). Best for: live pair-programming interviews where you want to see how a candidate uses AI, not just whether.

Which tools work in ordinary Zoom, Teams or Meet interviews?

Sherlock

Sherlock calls itself an interview proctoring agent. It joins Zoom, Google Meet and Teams calls as a participant once you connect your calendar, updates invites with secure meeting links, and sends real-time alerts if it detects AI assistance, deepfakes or proxy candidates. Starter begins at $100 a month for 10 interviews; Growth is $1,500 for 300. It claims detection accuracy above 97%, a figure from its own testing (Sherlock). Best for: remote-first employers interviewing for any role. Weak spot: a visible extra participant, and no published information on false positives.

Pindrop Pulse for Meetings

Pindrop says Pulse for Meetings flags proxy candidates and identity swaps, detects deepfake video and synthetic voices in real time, compares voices across interview rounds, and integrates with Zoom, Microsoft Teams and Webex. It reports a false positive rate below 1% on its own audio dataset (Pindrop). Its hiring page is unusually candid: detection "does not flag ordinary AI-assisted job seeking," and "a flag should trigger a documented human review, not an automatic rejection" (Pindrop hiring). Best for: enterprises hiring remote staff with system access.

Reality Defender and GetReal

Reality Defender's hiring solution alerts interview hosts to signs of manipulated video or voice while the candidate is on screen and flags AI callers on phone screens; its API has a free tier of 50 audio or image scans a month (Reality Defender). GetReal says it continuously authenticates identity during video and voice calls and has a page for staffing and recruiting firms (GetReal). Neither names meeting platforms or prices. Best for: organizations where security, not recruiting, owns the decision.

Interview Signal

Our desktop app includes a private answer check: when an answer reads as AI-written or read out, it tells the interviewer and suggests a question to try. It never changes a score and never appears in the scorecard or submittal. Best for: recruiters and hiring managers who want a nudge to probe, inside the interview they are already running. Weak spots: it does not detect deepfakes, verify identity or record video, and phone screens must run through the computer.

Can you trust an AI detection flag?

Treat it as a lead, not a finding. Accuracy figures on vendor sites come from the vendors' own tests. Real candidates also do things that look odd to a detector: pausing to think, glancing at their own notes, using assistive technology, or answering in a second language. A candidate with a disability may need an accommodation that changes how the interview looks; see ADA accommodations in interviews.

A fair process for any flag:

1. Ask an unscripted follow-up on the candidate's own example: a number, a name, a decision they made and why.
2. Note the flag and the follow-up answer in the interview record.
3. If doubt remains, run a second, short interview with a different interviewer, camera on.
4. Never reject on a flag alone. Our guide to handling a candidate who lies covers what to say.

What are the privacy and legal risks?

Identity matching and voice analysis can involve biometric data. Illinois' Biometric Information Privacy Act defines a biometric identifier to include a "voiceprint, or scan of hand or face geometry" and requires written notice of the purpose and retention period plus a written release before collection (740 ILCS 14, checked October 2026). Read our BIPA guide for hiring tools, ask each vendor exactly what it collects, and tell candidates before the interview. This is not legal advice.

How to choose a tool to detect AI use

• Technical roles with take-home or timed tests: use the integrity features of your assessment platform first.
• Remote roles with access to systems or money: add deepfake and identity checks, with a human review step.
• Everyone else: better questions catch more than software. Ask for specifics, follow up on the candidate's own examples, and verify employment history at the offer stage.