The Problem Changed in 18 Months
In early 2025, a handful of candidates were quietly pasting interview prompts into ChatGPT. By mid-2026, every smartphone ships with a generative AI assistant built in. Gartner (2025) estimates that one in five candidates now uses AI tools somewhere in the hiring process. That includes resume writing, cover letters, and yes, interview answers.
The instinct is to panic and pile on surveillance. Proctoring, screen monitoring, webcam tracking. That approach has two problems. First, it turns honest candidates into suspects. Cronofy (2024) reports that 42% of candidates abandon processes with too much friction. Surveillance adds friction. Second, determined cheaters find workarounds anyway. A second device next to the laptop. A friend reading answers off-camera. You cannot surveillance your way to trust.
The better question is not "how do we catch cheaters?" It is: how do we design interviews where cheating does not help?
Prevention Over Detection
The core insight is simple. ChatGPT can answer "what are the principles of good customer service?" but it cannot answer "tell me about the last time you handled an angry customer and what you would do differently." The first question tests knowledge that can be looked up in seconds. The second tests lived experience that cannot be generated on the fly.
Template design is the first and most effective line of defense. An interview full of googlable questions is easy to cheat on, no matter how many monitoring tools you bolt on. An interview full of experience-based, specific, contextual questions makes outside help nearly irrelevant.
Five Design Principles That Make Cheating Irrelevant
1. Ask about specific lived experience, not general knowledge
Replace "What is your approach to conflict resolution?" with "Describe the most recent conflict you had with a colleague. What happened, and what would you do differently?" The first question has a textbook answer on page one of any search engine. The second requires the candidate to draw from their own life. AI can generate a plausible fictional story, but the specificity of real experience is hard to fake and easy to probe in the next round.
2. Use video for the questions that matter most
Written-response questions are the easiest to cheat on. Copy, paste, done. Video responses are harder because the candidate has to speak the answer out loud, on camera, in their own words. A candidate reading a generated script on camera sounds different from someone speaking from experience. The pacing, the pauses, the way they return to a specific memory. Reserve your two or three most important questions for video format.
3. Set time limits that reward preparation, not searching
If a candidate has 10 minutes of prep time and 3 minutes to record, they have enough time to collect their thoughts but not enough to run an AI-assisted research session, edit the output, rehearse it, and deliver it naturally. The right window for most questions is 30 to 90 seconds of preparation and 1 to 3 minutes of answer time. Tight enough to favor people who know their own story. Loose enough that nervous candidates do not freeze.
4. Mix question types in a single interview
If every question is the same format, a cheater builds one workflow and repeats it. Mix video answers with ranking questions, code challenges, written responses, and multiple choice. Each type requires a different approach, which makes the effort compound. A candidate who can paste a written answer may not be able to cheat on a live video answer and a timed code challenge in the same session. Schmidt & Hunter (1998) showed that combining structured interviews with work sample tests outperforms either alone. The same principle applies: mixed evidence is harder to fabricate and more predictive of real performance.
5. Include one "react to this" scenario
Show the candidate a real situation and ask them to respond on the spot. For a customer support role: "This customer just sent this email. Record your reply." For a sales role: "A prospect raises this objection. How do you handle it?" For an engineering role: "Here is a short code snippet with a bug. Walk us through how you would find and fix it." The scenario does not need to be complicated. It just needs to be specific enough that the candidate cannot have a pre-written answer ready. This is the single most effective cheat-resistant question type because it combines domain knowledge, communication, and real-time thinking in one prompt.
What Not to Do
Do not pile on surveillance. Eye tracking, keystroke monitoring, secondary camera requirements, browser lockdowns. These make honest candidates feel watched and distrusted. The majority who answer honestly have a worse experience so you can catch the minority who do not. That math does not work. Design for the honest candidates.
Do not make the interview longer to compensate. Adding more questions "just in case" increases dropout without increasing signal. Cronofy (2024) is clear: friction kills completion. Five well-designed questions beat nine generic ones for both cheat resistance and candidate experience.
Do not assume everyone is cheating. The hiring process is a relationship that starts before the contract. If the first impression is "we do not trust you," candidates with options will walk.
The Template That Works
The best cheat-proof interview is one the candidate does not even recognize as cheat-proof. The questions feel natural and fair. The time limits feel reasonable. The mix of formats feels like a thorough evaluation, not a surveillance exercise. Honest candidates finish feeling good about the process. Cheaters find that their tools do not help them much.
That balance comes from template design, not from technology bolted on after the fact. Fifteen minutes building a template with experience-based questions, mixed formats, and one scenario question gives you a more cheat-resistant process than any amount of proctoring software.
Start a free trial of StormInterview and build your first cheat-resistant template this week. Five questions, three formats, one scenario. The candidates who do well on that template are the ones worth meeting in person.