Interviewing
What a Virtual Camera or AI Avatar Sets Off in a Video Interview
You can run one, and it works against you. An avatar or virtual camera over your own face is not the identity fraud federal alerts describe. It produces the tells hiring teams have been told to watch for: lip movement out of step with the audio, a fake background, a pause that lands wrong. What it invites is a closer look, such as an identity check or a second call with the camera on. If the real problem is being on camera, ask about that directly.
The takeAlmost everything published about avatars in hiring is written for employers buying detection, and it folds two unlike things into one category. One is somebody substituting a different person for the job. The other is somebody who does not want their face recorded, which is a normal request with a normal answer. Treating the second as a technical problem to solve quietly is what makes it look like the first.
Where Olive fits
Open a role and see what the work shows
An Olive assessment is not an interview and has no camera in it at all: you work a real assignment openly with an AI assistant, think aloud or type, and a person writes or edits the findings. The candidate is granted the same report the employer gets, free.
Assess your shortlistWhat Does an Avatar Do to Your Interview?
An avatar puts a rendered layer between your face and the person watching, and that layer is what gets noticed. The FBI's June 2022 announcement describes what complainants reported from online interviews: lip movement that did not fully coordinate with the audio, and coughs that did not line up with what was on screen 2. A rendered face can produce the same artifacts even when the person behind it is exactly who they claim to be.
Pindrop's Chief People Officer wrote up an applicant the company calls Ivan X in an April 2025 post. On the video call his expressions ran slightly out of sync with his words, his voice dropped out and misaligned with his lip movements, and he paused unnaturally at an unexpected technical question 3. That account is also an advertisement: Pindrop published it to launch a deepfake detection product for video meetings, nothing in it was independently verified, and one incident is not a rate.
Nobody publishes how often a virtual feed gets noticed.
What is published is the list the interviewer has in mind, and a rendered face sits in the middle of it.
Mandiant's September 2024 guidance on impostor hiring names fake backgrounds during interviews and reluctance to engage in video communication among the behaviours it saw in its own investigations 4. Those are one vendor's recommendations drawn from its own incident work, not a measurement, and every behaviour on the list has ordinary causes as well.
A slow connection produces some of the same picture.
The two things this question runs together are worth pulling apart. Putting a filter or an avatar over your own face, under your own name, for a job you intend to do yourself, is not the identity substitution those alerts describe. It still produces the picture that prompted them.
An avatar adds nothing to the answers you give.
Why Are Employers Asking for ID Now?
Because a run of stolen-identity hiring cases moved identity out of the interview and into a step of its own. In its 23 January 2025 alert the FBI said North Korean IT workers have been observed using artificial intelligence and face-swapping technology during video job interviews to obscure their identities 1. Its recommendations to employers include identity verification during interviewing, at onboarding, and throughout the employment of any remote worker 1.
Those are recommendations to employers, not a description of what any particular employer does, and the alert publishes no count of incidents. KnowBe4's own postmortem shows why the check moved outside the interview. Its HR team ran four video interviews on separate occasions and confirmed the person matched the photo on the application 5. The photo had been enhanced by AI from stock photography, and the background check came back clear because the stolen identity belonged to a real person 5.
In that one company's account of one incident, passing four video interviews and being who you say you are had stopped being the same fact.
So the ask arrives on its own. Greenhouse's Real Talent, launched 3 June 2025, verifies identity through CLEAR: a government ID and a selfie once inside the candidate's own Greenhouse account, then a selfie for later checks 6. Interviewers and hiring managers do not see either image, Greenhouse says, but are notified why a check failed 6. That is a vendor describing its own product rather than an audit of it, and it reaches only employers who switched the feature on.
The check is aimed at somebody pretending to be a different person.
Employers get told something on this ground worth reading from the other chair. The advice in what to do when the finalist may not be who did the take-home is to stop studying the video for tells and to walk every finalist through their own submission instead.
That advice exists because the video is a weak instrument here. Four interviews missed an impostor in the case above, and the artifacts a video does show have an innocent version.
Ask for a Format Without a Camera
If the problem is being on camera at all, say that to the recruiter instead of solving it at the video driver. Some interview formats have no camera in them. Sapia.ai's candidate explainer describes its chat interview as five to seven text-based questions, untimed, with no video recording and no camera at all 7. That is one vendor describing one product, so it tells you the format exists, not that the employer in front of you offers it.
Asking which formats are available is an ordinary question. Ask before you accept the slot, while a change is still easy to make.
If a disability is behind the request, it has a name and a route. The EEOC's 2022 technical assistance on AI and software assessments, which is not law and was taken off eeoc.gov in January 2025, said a request needs no particular words: a person does not have to mention the ADA or use the phrase reasonable accommodation for the request to count 9. The employer may still ask for supporting documentation where the need is not obvious.
What an assessment may record of your screen or your face covers what the consent screen is actually asking for, and what happens if you decline it.
The recording has its own route, written into state law. For a position based in Illinois, the Artificial Intelligence Video Interview Act, in force since 2020, gives you a deletion right on request 8. The employer then has 30 days, and must instruct everyone who received copies, the AI vendor included, to delete theirs and their backups too 8. The right is request-triggered rather than automatic, and it covers the videos themselves, not the notes or the ratings drawn from them. The Act names no penalty and no enforcement agency for a refusal.
It applies only where an employer asked for a recorded video interview and then ran an AI analysis of it.
How long a company keeps your AI interview video has the wording to send, and the limits on what a deletion request actually reaches.
Every one of these asks goes on the record under your own name.
What Happens If an Identity Check Flags You?
On one platform that publishes its rule, a failed check is a reason to look closer rather than a rejection. Greenhouse states that a failed identity verification does not automatically disqualify a candidate and that it indicates the need to investigate further 6. It also says its matching is built to handle name changes from marriage or immigration 6. What any individual employer does with the flag is still its own decision.
Greenhouse publishes no figure for how many legitimate candidates fail a check, so nothing here says how common a false flag is.
So the move is a question rather than a defence. Ask the recruiter what failed and what the next step is, and offer to redo the check or to join a short call with the camera on.
A repeat call with the camera on is one version of that next step.
There is one thing this whole process is trying to settle, and a camera is a weak way to settle it. Whoever did the work can talk about the work: the choice they backed out of, the thing that broke, the part that took two days longer than planned.
That holds up on a bad connection, in a borrowed room, with a face you would rather not have on file.
Common questions
Is an AI avatar the same thing as the deepfake candidates in the news?
No. The cases behind those alerts involve someone applying under another person's identity. Running an avatar over your own face, under your own name, is not that. The problem is that the feed can look the same either way. The published tells do not separate them: lip movement out of step with the audio, a fake background, a lag on an unexpected question.
Can an interviewer tell that I am using a virtual camera?
Nobody publishes a rate, so how often a virtual feed is noticed is unknown. What is published is the list of things interviewers are told to notice, and a rendered face produces several of them. The FBI's 2022 announcement describes lip movement that did not fully coordinate with the audio and audible actions that did not match the picture. Mandiant's guidance names fake backgrounds and reluctance to appear on video. Honest candidates on poor connections show some of the same signs, which is a real cost of that list rather than a loophole in it.
What if I just do not want to be on camera?
Ask the recruiter. Some formats are text-based with no camera at all, and asking whether a live call can replace a recorded video is an ordinary request. If a disability is behind the request, an accommodation route exists, and the EEOC's 2022 technical assistance, since taken off its site and not law, said the request needs no particular words and no mention of the ADA. Either way the ask goes on the record under your name, which is what separates it from a workaround.
What does an identity check actually involve?
On Greenhouse's Real Talent it is a government ID and a selfie, once, inside your own account, then a selfie for later checks. Greenhouse says interviewers and hiring managers do not see either image but are notified why a check failed, and that a failed check does not automatically disqualify a candidate. That is the vendor's description of its own product, not an audit, and it applies only to employers who turned the feature on. Other platforms and employers do it differently.
Can I get the video deleted afterwards?
For a position based in Illinois, the Artificial Intelligence Video Interview Act gives you a deletion right on request. The employer then has 30 days, and must instruct everyone who received copies, the AI vendor included, to delete theirs and their backups. It is request-triggered, so nothing happens until you ask. It covers the videos, not the notes or ratings drawn from them, names no penalty for a refusal, and reaches employers who asked for a recorded interview and ran an AI analysis of it.
References
- 1. North Korean IT Workers Conducting Data Extortion (Alert I-012325-PSA) ic3.gov Supports the statement that the FBI reported face-swapping technology used during video job interviews to obscure identities, and that its recommendations to employers include identity verification during interviewing, onboarding and throughout a remote worker's employment.
- 2. Deepfakes and Stolen PII Utilized to Apply for Remote Work Positions (Alert I-062822-PSA) ic3.gov Supports the described tells: lip movement that did not coordinate with the audio and audible actions that did not match what was shown on screen.
- 3. Think You Won't Be Targeted by Deepfake Candidates? Think Again. pindrop.com Supports the Ivan X account of expressions out of sync with words, audio dropping out or misaligned with lip movements, and an unnatural pause at an unexpected technical question.
- 4. Staying a Step Ahead: Mitigating the DPRK IT Worker Threat cloud.google.com Supports the statement that fake backgrounds during interviews and reluctance to engage in video communication appear among the behaviours Mandiant reported from its own investigations.
- 5. How a North Korean Fake IT Worker Tried to Infiltrate Us blog.knowbe4.com Supports the four video interviews that confirmed a match to an AI-enhanced application photo, and the background check that came back clear because the stolen identity belonged to a real person.
- 6. Making hiring safer and more human: An update on Real Talent greenhouse.com Supports the ID-and-selfie verification through CLEAR, that interviewers and hiring managers do not see the images and are notified only why a check failed, that a failed check does not automatically disqualify a candidate, and that matching is built to handle name changes.
- 7. AI Chat Interview Guide for Candidates: How It Works, Tips & FAQs sapia.ai Supports the existence of a camera-free interview format: five to seven text-based questions, untimed, with no video recording and no camera.
- 8. 820 ILCS 42, Artificial Intelligence Video Interview Act, Sec. 15 (Destruction of videos) ilga.gov Supports the Illinois deletion right: on request the employer must delete the interviews within 30 days and instruct everyone who received copies, backups included, to do the same.
- 9. The Americans with Disabilities Act and the Use of Software, Algorithms, and Artificial Intelligence to Assess Job Applicants and Employees web.archive.org Supports the statement that an accommodation request needs no particular words and no mention of the ADA, and the accompanying limit that an employer may still ask for supporting documentation where the need is not obvious.
9 sources, numbered by first appearance. How Olive sources claims
General guidance for hiring teams. What works at one company and one volume may not transfer to yours.
Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.