Taking an AI voice screen: what candidates should know

What is an AI voice screen?

An AI voice screen is a first-round interview in which a computer voice asks a set list of questions and records your spoken answers. You usually take it from a link, at a time you choose.

It replaces, or comes before, the first phone call with a recruiter. The questions are usually about your experience, why you want the role, and practical details such as when you could start. The interviewer may ask a short follow-up if an answer is brief. The invitation or the first page should tell you how long it takes, and it is fair to ask if it doesn't.

What is scored, and who decides?

Usually the content of your answers, rated against a rubric the employer wrote. Who makes the decision depends on the employer, so ask.

A rubric describes what a strong answer and a weak answer to each question contain. In Emoti's screens, each answer is rated on a four-point scale against that rubric, with a one-line summary, and a recruiter reads the results and decides whether you move on. Emoti doesn't reject anyone on its own. The camera, if it is used at all, never affects the score, and the screen doesn't try to read your emotions from your face or your voice.

Other tools work differently, and some laws give you rights here. In the EU, AI systems that infer emotions in the workplace are prohibited, and the European Commission's guidelines treat hiring as part of the workplace for that rule. The EU's GDPR also gives you the right not to be subject to a decision based solely on automated processing that significantly affects you, with limited exceptions. In New York City, employers that use an automated tool to assess candidates must tell candidates who live in the city in advance, including which qualifications and characteristics the tool assesses.

How should you prepare?

Prepare as you would for any structured interview: know your examples, practice out loud, and set up somewhere quiet.

1. Read the job post twice and list the three or four things it asks for most.
2. Pick an example for each. Know the situation, what you did, and what happened as a result.
3. Say your examples out loud and time them. Some screens, including Emoti's, show a time guide for each question.
4. Check the practical things: a quiet room, a charged device, headphones with a microphone, and a stable connection.
5. Choose your language. Some screens offer several languages. Pick the one you answer best in.
6. Read the first page carefully. It should say how long the screen takes, whether the camera is used, and who reviews your answers.

Expect questions in a few familiar areas, each with a simple shape for a good answer:

During the screen, take a breath before each answer. In Emoti's screens you can ask to hear a question again and pause between questions. If you are interrupted or cut off, keep going and tell the recruiter afterward.

How do you ask for an adjustment or a human alternative?

Write to the recruiter before you start, say what you need, and ask for the answer in writing. You decide how much to share about why.

Hello, I've been invited to the AI voice screen for the data engineer role. I have a stammer, and timed spoken answers put me at a disadvantage. Could I do this first round as a call with a person, or answer the questions in writing instead? Thank you.

Common requests include extra time, the questions in writing as well as aloud, answering in writing, a live interview with a person, and a break partway through. Asking early gives the employer time to arrange it.

The law may be on your side. In the US, the Americans with Disabilities Act requires covered employers to make reasonable accommodations for qualified applicants with disabilities, unless that would cause undue hardship. In Great Britain, the Equality Act 2010 places a duty on employers to make reasonable adjustments. In New York City, an employer's notice about an automated tool must include instructions for requesting an accommodation. In Illinois, an employer that uses AI to analyze a recorded video interview must explain how it works and get your consent first. None of this is legal advice.

What happens to the recording?

That depends on the employer and on the law where you are. Before you start, ask who can listen, how long the recording is kept, and how to have it deleted.

Three questions cover it: "Who will listen to or read my answers?", "How long will you keep the recording?" and "How do I ask for it to be deleted?" In Illinois, if an employer used AI to analyze your recorded video interview, it must delete the video within 30 days of your request. In the EU, you can ask any organization for a copy of the personal data it holds about you (GDPR Article 15).

In Emoti's screens, the first page tells you how long the screen takes, whether the camera is required, optional or off, the language and voice you will hear, and that people at the company review every answer before any decision. It doesn't state how long recordings are kept, so ask the employer.

What if the screen goes wrong?

Tell the recruiter in writing the same day: what happened, at which question, and what you would like to do about it.

Dropped connections, a misheard question or a noisy background happen. A short, factual note such as "My connection dropped during question three, and my answer was cut off" gives the recruiter a reason to read your answers with that in mind, or to offer you another attempt.

If you think the screen misheard you, say that too. Speech recognition is less accurate for some speakers than for others: one study of five commercial systems found an average word error rate of 0.35 for Black speakers and 0.19 for white speakers (Koenecke et al., 2020). A person reviewing your answers can only allow for an error they know about.

Employers who use these tools well want to hear about problems, because a screen that fails a strong candidate costs them a hire. If you want to understand how one screen is designed, including where its gaps are, Emoti publishes ours in Designing first-round voice screens candidates can trust.