AI in Recruitment

How Candidates Cheat in AI Interviews (And How to Stop It)

How Candidates Cheat in AI Interviews (And How to Stop It)
Abhishek Vijayvergiya
March 6, 2026
•
6 minutes

What Hiring Teams Need to Know

Recruiters need to know whether interview answers reflect a candidate's own skills. Outside assistance can produce a convincing first response while leaving gaps that become apparent only when the candidate has to explain a decision or complete related work.

Fabric's research on 19,368 interviews found that 38.5% triggered cheating flags. The dataset covers interviews conducted on Fabric between July 2025 and January 2026, with a flag defined as a cheating-probability score above 40%.

This guide explains four forms of interview cheating from the hiring team's perspective: what evidence to review, which follow-ups help verify skills, and how to handle uncertain results.

Four Methods and the Recruiter's Response

Fabric's research reports the following breakdown of classified cheating cases: tab switching and second screens, 18%; voice-mode LLMs, 34%; dedicated cheating tools, 45%; and live help from another person, 3%.

MethodShare of classified cases in Fabric’s researchEvidence to reviewRecruiter response
Tab switching and second screens18%Available focus events, candidate answers and work sample.Check permitted references and ask the candidate to explain a decision.
Voice-mode LLMs34%Consistency of answers and ability to apply the stated reasoning.Ask a role-specific follow-up and assess the explanation.
Dedicated cheating tools45%Interview report, relevant answers and follow-up performance.Review flags and test an edge case or changed requirement.
Live help from another person3%Continuity of experience, demonstrated work and relevant recording.Clarify the candidate’s own responsibility and investigate the specific concern.

Method 1: Tab Switching and Second Screens (18% of Cases)

This category covers candidates consulting outside material during an assessment intended to measure independent work. The hiring team's task is to establish whether the assistance was permitted and whether the candidate understands the answer they provide.

Where an interview platform records focus changes, review those events in context. A candidate may need an allowed reference, a coding environment or an accessibility tool.

Ask the candidate to explain a choice in their answer and apply it to a slightly different requirement. For example, after they suggest a data structure, ask what changes when memory is limited. Record the explanation and compare it with the role's assessment criteria.

Method 2: Voice Mode LLMs (34% of Cases)

Voice-based AI assistance can provide answers without appearing in the shared workspace. This makes the quality of the candidate's reasoning more useful to review than the screen alone.

Listen for whether a response addresses the actual question and remains consistent with earlier details. Then ask a related follow-up that requires the candidate to explain their own choice.

A practical assessment can help resolve uncertainty. For a customer-support role, present a new customer constraint and ask the candidate to explain how they would respond. Evaluate the decision and reasoning using the same rubric you apply to other candidates.

Method 3: Dedicated Cheating Tools (45% of Cases)

Tools such as Cluely can supply answers through an overlay that is invisible to screen sharing. A recruiter using Zoom, Teams or Google Meet cannot treat a clean shared screen as proof that no such tool is involved.

Fabric can detect Cluely-assisted interviewing. In Fabric's documented test, the interview received an 85% cheating-probability score.

Review the interview evidence and test whether the candidate can extend their answer. Ask them to explain an implementation decision, work through an edge case or connect an answer to a project they personally delivered. If the explanation does not hold together, clarify the gap before drawing a conclusion.

Our Cluely detection guide for recruiters covers the meeting-platform comparison and Fabric's test in more detail.

Method 4: Live Help From Another Person (3% of Cases)

This category includes another person supplying answers or taking part of an assessment on the candidate's behalf. Recruiters should look for continuity between the person, their claimed experience and the work demonstrated across stages.

Use specific questions about the candidate's own responsibilities. Ask what they built, which decision they owned, what went wrong and what they changed. A follow-up work sample can then test the same skill directly.

If a recording raises a concern about another person's involvement, establish what happened and whether it was permitted. Document the evidence and give the candidate a chance to clarify.

Why an Explicit AI-Use Policy Matters

Candidates and interviewers can have different assumptions about whether AI, documentation or notes are allowed. Hiring teams should resolve that ambiguity before the assessment rather than leaving recruiters to interpret it afterwards.

Define what each stage measures. One exercise might test independent reasoning; another might assess how someone uses AI and checks its output. State the rules in the invitation and repeat them at the start. Apply the same rules to candidates completing the same assessment, with agreed accommodations recorded.

How Can Employers Reduce ChatGPT Use During Interviews?

For an interview that prohibits outside answers, combine a clear policy with questions that require explanation and application. Use role-specific work, follow-ups based on the candidate's actual response and a consistent scoring rubric.

Avoid relying on repeated question banks alone. If an engineer explains a design, ask how it changes under a relevant constraint. If a salesperson describes a discovery call, ask what they would do with a different customer priority. Assess the reasoning the candidate can demonstrate.

Use Fabric's interview recording, transcript, assessment and flagged evidence to support review.

What Hiring Teams Should Do With a Flag

Review the specific evidence and the candidate's answers before deciding what the flag means. Note the assessment rules, the concern and the follow-up that would resolve it.

Where the result is unclear, ask for an explanation or arrange a short, relevant reassessment. Record the outcome so the recruiter, hiring manager and client can understand the decision.

For the full workflow and a client-submission checklist, see seven steps to prevent interview cheating in remote hiring. For an introduction to interview formats, see how AI interviews work.

Frequently Asked Questions

What Does Fabric's Research Say About Interview Cheating?

Fabric's study covers 19,368 Fabric interviews from July 2025 to January 2026. It reports that 38.5% triggered cheating flags, using a cheating-probability threshold above 40%.

Where Do the 18%, 34%, 45% and 3% Figures Come From?

They come from Fabric's published research. The report attributes 18% of classified cheating cases to tab switching and second screens, 34% to voice-mode LLMs, 45% to dedicated cheating tools and 3% to live help.

Can Employers Tell Whether a Candidate Is Using ChatGPT?

An employer can investigate possible outside assistance through interview evidence and relevant follow-ups. Fabric's AI interviewer flags possible assistance for review.

How Should Companies Follow Up on a Flagged Interview?

Review the interview evidence against the assessment rules and ask the candidate to explain any gaps. A focused work sample or follow-up interview can help verify the required skills. Record the outcome for the hiring manager.

What Should Recruiters Include in a Client Handoff?

Include the role criteria, work completed, relevant interview evidence and the outcome of any recruiter review. State what has been verified and what still needs a follow-up, so the client can make an informed decision. Use the client-handoff checklist to prepare the submission.

Review the Evidence With Fabric

Try a Fabric interview or book a demo to see the interview workflow and reports your hiring team can review.

See Fabric in action

Try Fabric for one of your job posts

Bring a role you’re hiring for. We’ll show you how Fabric can help.

Open booking page