For candidates
Do AI interviews use facial recognition or analysis?
The facial-scan fear, answered plainly. What AI interviews actually measure today, why the biggest vendor dropped facial analysis in 2021, the laws that now restrict it, and what you can do about it.
Almost never, at least not the way the phrase suggests. Facial recognition means identifying who you are, and AI interview tools already know that from your application. The real fear is facial analysis, software reading your expression to score you. The largest vendor dropped it in 2021, several laws now restrict it, and most AI interviews today score your words.
It is one of the most common worries about recorded interviews, and a reasonable one. You sit alone, talk into a webcam, and picture an algorithm grading every blink and micro-expression. This page separates the two things that get blurred together, recognition and analysis, lays out what the tools measure now, walks through the law in plain terms, and ends with what you can actually do. If you want the broader version of this question, is your interview scored by AI covers how to tell which kind of interview you are in.
Recognition versus analysis: two different things
The single most useful move is to split the fear in two, because the words get used interchangeably and they mean very different things.
Facial recognition is identity. It matches a face to a name, the way your phone unlocks or a passport gate works. Hiring tools have almost no reason to do this. You applied with your name, your resume, and often your email and phone number. The system is not guessing who you are. So the dramatic version, a camera secretly identifying you, is essentially not what is happening in a job interview.
Facial analysis is inference. It does not ask who you are, it tries to read something from your face, expression, gaze direction, how often you smile, and turn that into a number that feeds a score. This is the real subject of the worry, and the thing the 2019 to 2021 headlines were about. It is also the thing that has largely been pulled back. When a candidate online says a tool “scans your face,” analysis is almost always what they mean, even when they say recognition.
Keeping these apart matters, because the honest answer to “do AI interviews use facial recognition” is basically no, and the honest answer to “do any of them analyze faces” is “a shrinking few, and you can usually find out and opt out.”
What AI interviews actually measure today
Strip away the alarm and most AI-assisted interviews do something fairly ordinary. The tool records your answer, transcribes the speech to text, and scores that text against the criteria the employer set for the role. The signal is your words, the examples you give, whether you actually answered the question. A human then reviews the shortlist. This is the common case, and it is the one what an AI interview means describes in full.
Some tools also analyze tone or pacing from the audio, how fast you speak, how much you pause. That is voice, not face, and it tends to be a minor input. A smaller and now uncommon set of tools historically analyzed the video itself for expression and eye contact. That visual layer is the one that drew the lawsuits, the press, and the regulation, and it is the one that has faded fastest.
So when you record into a webcam, the most likely thing being judged is the transcript of what you said. Not your face. That is why the practical advice on a recorded interview is always about the answer, not about managing your expression: lead with your point, name specifics, structure the story. The same things that read well to a person read well to a transcript.
The HireVue turning point in 2021
Most of the facial-scan reputation traces to one company. HireVue is the best-known name in the category and functions as shorthand for the whole format. Facial analysis was once one of its headline features and, in plenty of candidates’ minds, its defining one. One recruiter who used the tool wrote that even years ago “most if not all” companies kept that module switched off, which is worth knowing on its own. But the bigger fact is what the vendor did next.
In 2021, HireVue announced it had removed facial analysis from its assessments. The reason it gave was not only the public pressure, though there was plenty. It was that the visual signal added little once natural-language analysis of what a candidate actually said had matured. In other words, the face was not pulling its weight as a predictor, so it went.
That single change matters out of proportion to one company, because HireVue was the reference point for the fear. The feature that built the facial-scan anxiety has not been part of the largest vendor’s scoring for years. If your worry is specifically “this looks like a HireVue, is it reading my face,” the honest answer is that the thing you are picturing was retired in 2021. For the platform specifics, how to do well in a HireVue interview covers what the experience is like now.
What the law actually says
The legal picture is the strongest reassurance, because it does not rely on any one vendor’s good behavior. In several places, analyzing your face is no longer something an employer can do quietly.
Illinois, the Artificial Intelligence Video Interview Act. In effect since 2020, this law requires an employer who uses AI to analyze a recorded video interview to do three things before you record: tell you AI is being used, explain in plain language how it works and what it evaluates, and get your consent. It also gives you a deletion right, on request the employer must destroy the video and its copies within 30 days. That 30-day rule is concrete and worth remembering.
Illinois, BIPA. Separately, the Biometric Information Privacy Act governs face and other biometric data in Illinois, requires consent before such data is collected, and carries real financial penalties. It is the law most often behind the headline settlements involving face data, and it is a major reason vendors grew cautious about visual analysis.
Beyond Illinois. The direction of travel is the same elsewhere. New York City requires bias audits of automated hiring tools under Local Law 144. Colorado has passed a broader AI-in-hiring law that phases in over the next couple of years. Other states and countries are moving similarly. You do not need to memorize any of this. The takeaway is that “is anything analyzing my face, and how” is a fair question, and in a growing number of places the employer is legally obligated to answer it before you start. The full version, including who can see the recording and how long it is kept, is in is it legal for an employer to record your interview.
So what is actually left
Put the pieces together and the residual risk is small and, importantly, knowable.
The likeliest case by far is that a person watches your recording, or that AI scores the transcript of what you said. A modest case is that audio tone or pacing is one input among several. The case people fear most, software grading your expression and eye movement, is now the rare exception. It peaked around 2019 to 2021, lost its biggest proponent when HireVue dropped it, and runs into consent and biometric law in a widening set of places.
None of this means the format is above criticism. Candidates have raised fair concerns about bias in any automated scoring, particularly for neurodiverse and international applicants, and those concerns are part of why the visual layer retreated and the law tightened. Are one-way interviews fair takes that question on directly. The point here is narrower and more practical: the specific thing called facial recognition is almost never happening, and facial analysis is now the exception you can ask about, not the default you should assume.
What you can actually do
If the question is still nagging you before a specific interview, you have real options, not just reassurance.
- Ask the recruiter directly. A short email is normal and most employers answer it: “Could you tell me what the assessment evaluates, whether it analyzes anything visual, and whether a person also reviews my answers?” In Illinois and a growing list of places, they have to tell you before you record anyway.
- Look up the platform. The invitation almost always names the tool. Most vendors publish what their scoring does, and you can usually confirm in a minute or two whether visual analysis is even on the table.
- Know your deletion and consent rights. Where laws like Illinois apply, you can ask for the recording to be deleted, within 30 days there, and consent is required before any AI analysis. If you have a disability that the format disadvantages, you can also request an accommodation, covered in video interview accommodations.
- Then prepare the answer, not the face. Once you know what the tool does, the highest-value preparation is the same as for any recorded interview. A clear point in the first ten seconds, specific examples, a clean and well-lit setup. That is what carries weight with a human reviewer and a transcript, which is what is almost certainly scoring you. How to prepare for an AI interview walks through it.
The fear of a machine reading your face is understandable, but it is largely a fear about a feature that has been retired by its biggest user and restricted by law. Treat the recorded, one-way interview as what it usually is, a consistent way to answer the same questions every other candidate gets, and put your energy where it actually moves the result.
If you want to confirm which kind of interview you are facing, start with is your interview scored by AI, and if it is a recorded round, how to prepare for an AI interview is the next step.