# Video interview cheating: what candidates try and how to stop it

> Video interview cheating happens, usually scripts or AI answers read aloud. See the warning signs and how to write questions that make faking pointless.

Source: https://deephire.com/blog/cheating-in-one-way-video-interviews
Last checked against the product: 2026-11-18

## Key takeaways

- Candidates can cheat in one-way video interviews, mostly by reading prepared or AI-written answers, or by getting help from someone off camera.
- Questions about the candidate's own past work are much harder to fake than generic ones like tell me about yourself.
- Time limits that fit the question and deliberate retake settings take away most of the payoff from cheating, especially when you also rotate questions.
- Treat warning signs as a reason to probe in the live round, not as proof, because nerves and disabilities can look similar.
- Put your rules in the invite and tell candidates you will ask follow-up questions about their answers later.

Yes, candidates can cheat in one-way video interviews, the same way they can read notes during a phone screen or get coached before an onsite. Most video interview cheating is low tech, such as reading a prepared script or getting help from someone off camera. Specific questions and a live conversation before any offer make most of it pointless, and sensible time limits handle a good part of the rest.

The format does open a few gaps that a live interview doesn't have. Nobody watches in real time and the candidate controls the room. An AI chat tool can also turn a generic question into a polished answer in seconds. If you are new to the format, start with [how a one-way video interview works](https://deephire.com/blog/what-is-a-one-way-video-interview).

## How do candidates cheat in one-way video interviews?

Most attempts fall into one of these:

1. Reading a script. The candidate writes answers in advance and reads them from a document next to the camera.
2. Reading AI-generated answers. The candidate types the question into a chatbot during prep time and reads the output aloud.
3. Off-camera help. Another person, such as a friend or a paid coach, sits out of frame and feeds lines, or the candidate gets texts during the answer.
4. Using a stand-in, where someone else records the interview for the candidate. This is less common, but remote roles and roles with skills questions are more exposed to it.
5. Sharing questions. Candidates who already interviewed post the questions online or pass them to friends applying for the same job.
6. Looking things up. For technical or knowledge questions, the candidate searches for the answer while recording.

Almost all of these depend on generic questions like "Tell me about yourself" or "What is your greatest weakness?" Those have well-known model answers, and a script or a chatbot can produce one instantly. A question about the candidate's own work history is much harder to fake, because nobody else knows the answer.

## What counts as cheating, and what is just preparation?

Before you try to catch anyone, decide what you actually object to. A lot of what looks like cheating is normal preparation that you would accept in any other interview.

| Behavior | Usually fine? | Why |
|---|---|---|
| Practicing answers before the interview | Yes | You want prepared candidates |
| Using retakes you allowed | Yes | You set the rule |
| Glancing at a few bullet-point notes | Usually | Same as notes in a phone screen |
| Reading a full script word for word | Depends on the role | Hides how they communicate |
| Reading answers from an AI tool while recording | Usually not | The answer is not theirs |
| Someone else feeding answers or recording | No | You are not assessing the candidate |

The middle rows are judgment calls, and the job decides them. In a customer service role where the person will talk to callers all day, a read-aloud answer hides the exact skill you are screening for. In a back-office data role, notes may not matter much. Write your rule down and tell candidates up front. Then score everyone against it.

## How to spot cheating in a recorded answer

There is no single tell. Look for a pattern across several answers instead of reacting to one odd moment.

These signs are worth a closer look:

- Eyes tracking steadily left to right across a fixed point, the way they do when reading lines of text
- A flat, even delivery that sounds nothing like the candidate's tone in unscripted moments, such as their greeting
- Answers that are polished but generic, with no names of tools, team sizes, customers, or outcomes
- Several candidates using nearly the same unusual phrasing for the same question
- Pauses where the candidate looks off screen and then resumes with a complete, formal sentence
- A second voice or typing sounds during the answer
- The candidate reacting to someone out of frame
- A face or voice that does not match a later live call

These signs often mislead reviewers:

- Looking away while thinking, which many people do naturally
- Looking slightly off center, often because the camera sits above a second monitor or the candidate is watching their own preview
- A stiff or rehearsed tone, which nerves and camera shyness cause too
- Unusual eye movement or speech patterns, which can come from a disability or neurodivergence and say nothing about honesty

> Treat warning signs as a reason to probe in the next round, not as proof.

Shared scripts are easiest to catch when you watch every candidate's answer to the same question back to back. DeepHire lets you compare answers question by question, which makes copied phrasing much easier to notice than reviewing one candidate at a time.

## How to design a one-way interview that is hard to game

Detection only gets you so far. The better goal is a screen where the most useful preparation a candidate can do is think about their own experience.

### 1. Ask about their own work, not general opinions

Swap generic prompts for questions that require specific memories. Instead of asking how they handle difficult customers, ask about the last time a customer was upset with them and what they did next. A chatbot can write a convincing answer to the first version. It has no way of knowing what happened at their last job. Our list of [pre-screening interview questions](https://deephire.com/blog/pre-screening-interview-questions) has plenty of examples you can adapt.

### 2. Build a follow-up into the question

Add a second part that asks for detail, such as what the result was or who else was involved. A true story usually holds up when you ask for the outcome or the people in it. An invented one tends to go vague at exactly that point.

### 3. Set time limits that fit the question

A short prep window and an answer limit of one to three minutes leave little room to run the question through a chatbot and rehearse the output. Don't squeeze so hard that nervous candidates freeze. For most screening questions, a minute or two to answer is plenty.

### 4. Decide on retakes deliberately

Retakes help nervous candidates show their best, and on a question about their own experience they don't give a cheater much. Unlimited retakes on a question with well-known answers are different, because they let a candidate polish a script. Consider allowing retakes on the warm-up question and limiting them on the one or two questions that matter most. DeepHire lets you set how many retakes candidates get, and the default is unlimited, so change it if your roles call for it.

### 5. Rotate questions before they spread

If questions leak, rotate them. Keep two or three versions of each key question and swap them every hiring cycle, or every few weeks for high-volume roles.

### 6. Add one practical, role-specific task

Where it fits, ask the candidate to do something on camera. They could explain a process to a new hire or respond to a sample customer complaint written into the question. For shift-based roles, you can ask them to walk through how they would handle the first hour of a shift. Tasks like these are hard to outsource in real time.

### 7. Tell candidates the rules

Most people follow clear instructions, so put your rules in the invite. Wording like this works:

- Short notes are fine. Please do not read a full script.
- Please answer on your own, without help from another person or an AI tool while recording.
- We will ask follow-up questions about your answers in the next interview.

The third line does the most work, because a candidate who expects the live interviewer to dig into their story has little reason to invent one. If you use AI to summarize or score answers on your side, be open about that too. Our post on [AI video interview disclosure](https://deephire.com/blog/ai-video-interview-disclosure) covers how.

## What to do when you suspect a candidate cheated

Don't reject on a hunch alone, and don't accuse anyone in writing. Work through these steps instead.

1. Note what you saw and the timestamp in your shared comments, in neutral language. Write something like "answer to Q3 appears read, little specific detail" instead of "cheated."
2. Score the answer on its content against your rubric. A read-aloud generic answer usually scores low anyway because it lacks specifics.
3. If the candidate otherwise looks strong, move them forward and plan targeted follow-ups for the live round. Ask them to expand on the exact story from their recording.
4. For roles where identity matters, confirm identity at the live interview or during onboarding using your normal process.
5. Apply the same rule to every candidate, since inconsistent treatment is unfair and creates risk. Rules on identity checks, recording, and AI use in hiring vary by place, so check with your legal team before adding new verification steps.

## How do you keep the screen fair for honest candidates?

Worrying too much about cheating can make the screen worse for everyone else. Heavy suspicion leads to tight time limits and zero retakes, and sometimes to rejections based on where someone's eyes went. That screens out honest candidates who are nervous or recording on a phone in a busy house, while a determined cheater just adapts.

Keep the bar on whether the candidate can describe real experience clearly and specifically. Score with a written rubric that is the same for everyone, so one reviewer's feeling about eye contact doesn't decide the outcome. If you need a starting point, our [hourly candidate scorecard](https://deephire.com/blog/hourly-candidate-scorecard) template adapts easily to salaried roles.

The recorded interview is a first screen whose job is to tell you who deserves a conversation. In that conversation, ask the candidate what happened next in the story they recorded, and you will find out whether a polished answer had anything behind it.

## Frequently asked questions

### Can recruiters tell if you are reading during a video interview?

Often, yes. Steady side-to-side eye movement and a flat delivery are common giveaways, especially when the answers are polished but generic. None of these is proof on its own, so a careful recruiter follows up in a live interview instead of rejecting on that basis.

### Is using an AI chatbot during a one-way video interview cheating?

Most employers treat reading AI-generated answers while recording as cheating, because the answer is not the candidate's own. Using AI to practice before the interview is usually fine. State your rule in the interview invite so candidates know where the line is.

### Do one-way video interview platforms detect cheating automatically?

It depends on the tool. Many record the answers for a person to review, and any automated flags should be treated as prompts for a closer look, not verdicts. Specific questions and a live follow-up round are the more dependable defense.

### How do you stop candidates from sharing interview questions?

You can't fully stop it, so make shared questions less useful. Ask about the candidate's own experience, which no one else can answer for them, and rotate two or three versions of each key question over time.
