Interviewing
What an AI Interview Scores, and What It Does Not
No published source says what every AI interview scores. HireVue, one large vendor, states its AI reads only the words in an answer, not facial expression, tone, or body language. In the EU, inferring emotions in the workplace from face or voice has been prohibited since 2 February 2025, though the text does not settle whether a pre-hire interview counts. You do not have to guess: an Illinois role gets a legal disclosure, a New York City role gets a posted audit. Ask.
Where Olive fits
Open a role and see what the work shows
An Olive session is not an interview and nothing in it is scored from your voice or face: you work a real assignment with an AI assistant, think aloud or type, and a human reviewer writes what happened in six named findings. You are handed the identical report the employer reads.
Rank your shortlistWhat does the AI actually read?
On the one large platform that publishes an answer, an AI interview scores what you said, not how you looked saying it. HireVue describes a three-stage pipeline: speech to text first, then a language model reading the transcript, then a score benchmarked against expert human raters. The same statement names the limits directly: no video analysis, and no scoring of facial expression, body language, background, or tone of voice 1.
That description answers the fear behind the most common version of this question, some variant of "is it reading my face." For that platform, the vendor's own answer is no. Separately, and for a wider set of tools, the EU AI Act prohibits inferring a person's emotions in the workplace from biometric data, a ban that has applied since 2 February 2025 2. The Act says "workplace and education institutions" and does not settle whether a pre-hire interview with an outside candidate sits inside that.
Neither fact proves every vendor scores identically, and this article is not claiming that. No independent audit of any AI interview vendor's scoring is cited here, because none was available to check. What the two facts do show is that the claim most prep advice still repeats, that a camera is silently grading your expression, is not what the vendor with the most detailed public statement says its product does, and is prohibited outright in one large jurisdiction. Read a published vendor statement as exactly that, a vendor describing its own product, and hold it against the law rather than picking whichever one is more comforting.
The transcript itself is not immune to error, either. A speech recognizer produces it before any scoring happens, and a recognizer can misfire on background noise, a fast talker, or an unfamiliar accent well before your words ever reach a rubric. That is a real, separate risk worth asking a recruiter about, and it is a transcription question rather than a face-reading one, which is exactly why the two get confused.
How do you find out what your interview specifically scores?
Two disclosures exist in US law right now, and both name the employer's obligation rather than leave it to a candidate's guess. If the role is based in Illinois, the Artificial Intelligence Video Interview Act, in force since 1 January 2020, requires the employer to explain, before your interview, how the AI works and what general types of characteristics it evaluates, and to get your consent first 3.
If the role is in New York City, Local Law 144, in effect since 1 January 2023, requires the employer to have run a bias audit within the past year, posted a public summary of it, and given you notice at least 10 business days before the tool is used 4. Neither law tells you the scoring weights or the exact rubric. The Illinois Act names no penalty and no enforcement agency at all, while Local Law 144 is enforced by the city with civil penalties, so read the first as a right to ask and the second as a duty someone can be fined over.
Where neither law applies, one sentence to a recruiter is the shortest route: what does this interview score, and is any part of it automated? A recruiter who cannot answer usually has not been handed the vendor's documentation, which is itself information. The employer side of exactly this question walks through what to ask a vendor to verify a bias-testing claim rather than take it on faith, and a shorter version of the same question works from your side of the desk.
You do not need a lawyer for this. The Illinois disclosure and the New York City audit summary are both meant to be plain-language, and a recruiter forwarding you a link to either one is doing the minimum the law actually asks of them. If what comes back is vague marketing language rather than a description of what gets measured, that gap is itself worth noting before you decide how much weight to put on the round. Naming the document you are asking for, the AIVIA disclosure or the Local Law 144 audit summary, tends to get a faster and more specific answer than asking generally whether the process is fair.
What survives no matter which vendor you meet?
A rubric scores what you said, and a rubric does not know the difference between a smooth delivery and a flat one when the content underneath is the same. That is not a promise about any specific outcome; it is what a rubric-based system is built to be blind to, by design, because a written scale reads the same words whichever way they were spoken.
A specific answer with a real decision and a real number in it gives a transcript something concrete to grade. A generic answer gives it nothing to distinguish from the next generic answer. Spend your prep there instead of on posture. Pull the actual competencies out of the job posting, write one real example per competency, and say the specific thing rather than the impressive-sounding thing. The prep guide for the recorded round walks through this in an hour, competency by competency, and the same logic holds whether the round in front of you is recorded, live, or scored off a human interviewer's transcript.
One more habit travels across every format described here. Whatever you are told the interview scores, write it down afterward, in your own words, along with which questions came up and roughly how the round went. The recording, the transcript and the score sit in the employer's and the vendor's systems rather than yours, and getting a copy runs through a specific route rather than a general right: one federal rule can put a vendor's report about you in your hands before a decision is final. Your own note is the record you hold without asking anyone for it.
Common questions
Does an AI interview score my accent or how fluently I speak?
On a system that scores transcript content only, an accent does not change the words on the page, so it should not change the score directly. A transcription error introduced before the words ever reach the scorer is a separate, real risk, and it is worth asking who reviews an answer that looks garbled in the transcript.
Can I ask to see the rubric before my interview?
Asking is reasonable, and some employers will share the competencies even if not the exact scoring weights. Framing it as understanding what to prepare, rather than as a challenge to the process, tends to get a fuller answer. A recruiter who declines entirely is telling you something too.
Does the EU ban on emotion inference mean my video interview isn't recorded?
No. The recording still exists in most systems, and a recruiter may still watch it. Article 5(1)(f) of the EU AI Act bans using AI to infer emotions from biometric data in the workplace and in education institutions, in force since 2 February 2025; it does not require a tool to stop capturing video, and it does not reach a written transcript's content. Whether a pre-hire interview with an outside candidate falls inside "the workplace" is not settled by the text of the Act.
What if I'm applying from outside Illinois or New York City?
Neither law reaches you directly, but the underlying question is still fair to ask a recruiter anywhere, and a vendor's own published explainability statement, where one exists, applies regardless of where you are sitting for the interview.
Is there any evidence the AI is actually reading my face despite what a vendor says?
A vendor's own statement is not an independent audit, and this article treats it as a vendor claim rather than proof. No independent audit of any AI interview vendor's scoring is cited here, because none was available to check. What sits in a primary source is the EU's prohibition on inferring workplace emotions from face or voice, which has applied since 2 February 2025.
References
- 1. 2025 Explainability Statement hirevue.com Supports that HireVue's AI scores only the transcribed words of an answer and does not analyze video, facial expression, background or tone of voice.
- 2. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) publications.europa.eu Supports that the EU AI Act's Article 5(1)(f) ban on inferring workplace emotions from biometric data has applied since 2 February 2025.
- 3. Artificial Intelligence Video Interview Act, 820 ILCS 42 ilga.gov Supports that an Illinois-based employer using AI video analysis must explain how it works and what characteristics it evaluates, and obtain consent, before the interview.
- 4. Automated Employment Decision Tools: Frequently Asked Questions nyc.gov Supports that a NYC employer using an automated employment decision tool must have a bias audit posted and give candidates notice at least 10 business days beforehand.
4 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.