--- title: Proxy Interviews and the Deepfake Crackdown in India description: Interview fraud is rising and verification is tightening in India. What employers now check, and how honest candidates avoid looking suspicious. url: https://usegreenroom.app/blog/proxy-interviews-deepfake-crackdown-india last_updated: 2026-08-01 --- ← Back to blog India · Careers # Proxy interviews, deepfakes, and the verification crackdown August 1, 2026 · 12 min read ![Proxy interviews and deepfake crackdown in India — candidate guide cover from Greenroom, the AI mock interviewer](/assets/blog/proxy-interviews-deepfake-crackdown-india-hero.webp) He was not cheating. He had two monitors, a habit of glancing at the second one while thinking, a slightly laggy connection, and a flatmate who walked through the frame twice. At the end of the round the interviewer asked him to turn his camera to show the room, and the whole thing became briefly, unpleasantly awkward — not because he had done anything, but because 2026 is a year in which interviewers have learned to be suspicious. They have reasons. A widely-cited Greenhouse survey of just over four thousand respondents reported that **91% of hiring managers had encountered or suspected AI-assisted cheating in interviews**, and that **18% had caught a deepfake**. An EY analysis of more than a million pre-employment screenings across 90-plus organisations in India reported a significant rise in employment fraud — and, notably, more experienced professionals being caught than freshers. Screening vendors report finding fraudulent activity in a quarter to a third of sessions they flag as suspicious. These are vendor and consultancy figures rather than independent research, so treat the precise numbers as directional, but the trend is consistent across every source. The consequence for you, if you are honest — which is almost everyone reading this — is that **the process is getting more suspicious of you regardless**. This guide covers what is actually being checked and how not to trip over it. ## What is actually happening Three distinct things get bundled together as "interview fraud": - **Proxy interviewing.** Someone else attends the interview in your place, or sits off-camera feeding answers. This is the oldest version and remains the most common in India. There was a widely-reported case in which a candidate joined a large Indian IT services firm after a friend impersonated him in the video interview; his performance gave him away within two weeks and it ended in dismissal and a criminal complaint. - **AI copilots.** Software listening to the interview and generating answers on a second screen in real time. An entire product category exists for this. - **Deepfakes.** Real-time face and voice manipulation so the person on camera is not the person applying. Rarer, harder, and disproportionately associated with fully-remote roles and identity fraud rather than ordinary candidates. ## What employers now check ![Diagram of what interview verification now checks — identity at the start, camera and environment during the round, live follow-up questions, screen and audio signals, and post-offer background verification](/assets/blog/proxy-interviews-deepfake-crackdown-india-diagram.webp) Five checkpoints, most of them invisible. None of them are a problem if you are who you say you are — but two of them catch honest candidates by surprise. - **Identity at the start.** Government photo ID held up to the camera, matched against your face and your application. Increasingly routine, including at mid-size Indian companies. - **Camera on, face visible, whole time.** Refusing camera is now treated as a signal rather than a preference. - **Environment checks.** Being asked to pan the room, or to confirm you are alone. Uncomfortable, and becoming normal for high-stakes rounds. - **Live, unscripted follow-ups.** The cheapest and most effective test there is. A copilot can produce a fluent first answer; it cannot survive "you said X — why not Y, and what happened when you tried?" three times. This is why good interviewers have gone deeper rather than harder. - **Behavioural signals during the round.** Long uniform pauses before every answer, eyes tracking a second screen, typing sounds, a second voice, or answers that are suspiciously polished relative to your spoken register. - **Proctored assessments.** Tab-focus tracking, copy-paste detection and full-screen enforcement on online tests. - **Post-offer background verification.** Dates, designation, last salary and rehire eligibility. Our [background verification in India](/blog/background-verification-india) guide covers this in detail. ## How honest candidates get caught out This is the part worth your attention, because two of these catch genuinely honest people: - **Glancing at a second monitor.** If you think by looking away, say so early: "I'll be looking at my notes on the second screen occasionally." One sentence removes the ambiguity entirely. - **Notes.** Usually fine, occasionally not. Ask at the start whether notes are acceptable. Silence plus visible reading looks worse than asking. - **A borrowed or shared machine** whose account name does not match yours. - **A name mismatch** between your display name, your CV and your ID — extremely common with nicknames and initials, and it creates a moment of doubt you do not need. - **Your answers being much more polished than your speech.** If you memorised a script, the delivery mismatch is genuinely noticeable and reads as assistance. - **Poor connection causing lag.** Lag looks like delay, and delay looks like consulting something. - **Someone walking through frame.** Harmless; still noted. None of these are accusations. They are ambiguity, and ambiguity in 2026 costs you more than it used to. ## What to do about it, practically - **Show ID without being asked** if the invite mentions verification. It takes ten seconds and it closes the question. - **Match your names.** Display name, CV, and ID — the same version of your name in all three. - **Declare your setup in the first thirty seconds.** Notes, second monitor, someone else at home. Voluntary disclosure is treated completely differently from discovered behaviour. - **Use a plain background, alone, with the door shut.** Not for privacy — for the absence of anything to wonder about. - **Prefer wired or strong wifi.** Lag is the most common innocent behaviour that reads as guilty. - **Do not use a copilot.** Beyond the ethics, detection is improving quickly and the consequence is not a rejection — it is a blacklist, and in the reported Indian cases, sometimes a criminal complaint. - **Speak your own answers.** The most robust anti-suspicion strategy is fluency in your own material, which is a preparation problem rather than a technology one. **The core truth:** you cannot control how suspicious the process has become, but almost all of the risk to an honest candidate is ambiguity — a second screen, an unexplained pause, a mismatched name. Ten minutes of setup and one declarative sentence at the start removes nearly all of it. ## If you are wrongly accused It happens, and it is worth knowing the response. Stay calm and factual. Ask specifically what was observed rather than defending in general. Offer to demonstrate — show the room, share your screen, redo a question live with the camera on your hands. Then answer a technical question in depth, out loud, because genuine knowledge is very hard to fake under follow-up and is the strongest evidence available to you in the moment. If the round ends badly, email the recruiter the same day with a short, non-defensive account and an offer to redo the round under any conditions they like. Most companies would rather resolve it than lose a candidate they liked. ## What this means for how you prepare The crackdown has a genuinely useful side effect for honest candidates: **interviews are moving towards deeper follow-ups and away from answers that can be looked up.** That is a format that rewards people who actually know their own work and can talk about it, and penalises people who prepared by reading. Which means the highest-value preparation in 2026 is the least technological one — saying your own answers out loud until they hold up under three follow-ups. That is what our [how to explain your project in an interview](/blog/how-to-explain-your-project-in-an-interview) guide covers, and it is the entire premise of [Ari, the AI interviewer](/): a spoken round where the next question depends on what you just said. To be explicit about our own position, since this post is about integrity: Greenroom is practice **before** the interview and is deliberately not a copilot **during** one. Our [cheating signals](/blog/cheating-signals-and-integrity) page sets out how we think about this, and our [Greenroom vs Final Round AI](/blog/greenroom-vs-final-round-ai) comparison covers the difference from tools built for the other purpose. ## Where each option actually helps - **The recruiter** — ask directly what verification the process includes. They will tell you, and knowing removes the surprise. - **Your ID and documents, ready before the call** — a two-minute preparation that prevents a five-minute fumble. - **A wired connection or a tested hotspot** — the single highest-value fix, because lag is what most often reads as suspicious. - **ChatGPT** — fine for preparing content beforehand. Using it during a live interview is the specific thing being detected. - **Greenroom** — the spoken layer, before the interview. [Ari](/) pushes the follow-ups that a copilot cannot survive, which is exactly the direction interviews are moving. Honest tradeoff: it prepares you; it does not sit with you. Not every use of AI is prohibited now — our [can you use AI in a coding interview](/blog/can-you-use-ai-in-a-coding-interview) guide covers the companies that permit it and how they grade it. ## Frequently asked questions ### What is a proxy interview? A proxy interview is when someone other than the applicant attends the interview, either by impersonating them on camera or by sitting off-screen supplying answers. It is the oldest and still most common form of interview fraud in India, and it has led to dismissals and criminal complaints in reported cases — including a widely-covered incident where a candidate joined a large Indian IT services firm after a friend impersonated him and was found out within two weeks of starting. ### How do companies detect AI cheating in interviews? The most effective method is not technological: unscripted live follow-up questions, because a copilot can produce a fluent first answer but cannot survive being asked why not the alternative, and what happened when you tried, three times in a row. Alongside that, employers use identity checks at the start, camera-on requirements and occasional room checks, behavioural signals such as uniform pauses before every answer or eyes tracking a second screen, proctored assessments with tab-focus and copy-paste detection, and post-offer background verification. ### Is it OK to use notes in a video interview? Usually yes, but ask at the start rather than assuming. The problem is never the notes, it is the ambiguity — visible reading with no explanation looks like assistance, whereas saying in the first thirty seconds that you have notes on a second screen and will glance at them removes the question entirely. Voluntary disclosure is treated completely differently from behaviour that gets noticed and interpreted. ### How can an honest candidate avoid looking suspicious in an interview? Most of the risk is ambiguity rather than accusation. Match your display name, CV and ID to the same version of your name, keep the camera on with a plain background in a room where you are alone, declare notes or a second monitor in the first thirty seconds, use a wired or strong connection because lag reads as delay and delay reads as consulting something, and have your ID ready if the invite mentions verification. ### What should you do if you are wrongly accused of cheating in an interview? Stay factual and ask specifically what was observed rather than defending yourself in general terms. Offer to demonstrate — show the room, share your screen, or redo a question live — and then answer a technical question in real depth, because genuine knowledge is very hard to fake under follow-up and is your strongest available evidence. If the round ends badly, email the recruiter the same day with a short non-defensive account and an offer to redo it under any conditions they choose. ### Are AI interview copilots worth the risk? No. Detection is improving quickly, the consequence is typically not a simple rejection but removal from consideration permanently and in some reported Indian cases a criminal complaint, and the format of interviews is actively shifting towards deep unscripted follow-ups that copilots handle badly. The more durable strategy is fluency in your own material, which is a preparation problem rather than a technology one. Interviews are moving towards deeper follow-ups precisely because those are hard to fake — which rewards anyone who has actually rehearsed their own answers. [Greenroom](https://usegreenroom.app/) is practice before the round, never a copilot during it. Free to start. Curious how it works? See [how AI mock interviews work](/blog/ai-mock-interview).