What Is an AI Interviewer? Async and Conversational Interviews Explained

An AI interviewer is software that conducts a structured interview with a candidate without a recruiter present, then produces a transcript and an assessment against criteria you defined in advance. The candidate answers by video or voice, on their own schedule. The system records, transcribes, and scores.

The term covers two quite different things, and conflating them is where most confusion starts.

Two things called the same name

What it actually measures

An AI interviewer measures what a candidate said, matched against what you said you wanted. That is genuinely useful for a specific, narrow question: does this person have the experience the role requires, described in their own words rather than written by a CV service?

It is much weaker at the things interviews are often assumed to establish.

• Verbal fluency is not job performance. A candidate who interviews well in their second language and one who does the job well are different populations. In a market where a large share of candidates are interviewing in English as a second or third language, this gap is not hypothetical.
• Inferences about personality from voice or face are not reliable. Treat any product claiming to score confidence, enthusiasm or cultural fit from video with suspicion, and check what your legal obligations are before enabling it.
• Scores are not neutral. A score reflects the criteria you gave it. If the criteria encode a preference, the score will too, more consistently than a human would.

What Saudi employers need to know before switching it on

In practice that means three things for a hiring team.

• Keep a human in the decision, or collect explicit consent. Those are the two lawful paths, and the first is simpler. If a person makes the reject decision, the "solely automated" trigger never fires and the explicit-consent requirement never applies.
• Candidates are told. Disclose that the interview is automated, what is recorded, how long it is kept, and who sees it. Do this before the interview, not in a policy page nobody opens.
• There is a way out. Offer a route to a human conversation for candidates who ask. Some will, and refusing looks worse than the cost of accommodating them.

When it is worth it, and when it is not

Asynchronous interviewing earns its place when the ratio is brutal and the questions are genuinely uniform: a few hundred applicants for a handful of similar openings, where the alternative is that most candidates get no conversation at all. Used that way, it widens access rather than narrowing it.

It is a poor fit for senior or scarce roles. A candidate with three offers will not record answers into a browser tab, and asking them to signals how the rest of the process will feel. For those roles the bottleneck is not screening throughput, it is your ability to sell the job.

Questions worth asking a vendor

• What exactly does the score represent, and can I see the reasoning behind an individual score?
• Does it work in Arabic, including transcription of Gulf dialects, or only Modern Standard Arabic?
• Where is candidate video stored, in which country, and for how long?
• Can I turn off any inference drawn from face or tone of voice?
• What happens when a candidate asks for a human interview instead?
• Can a candidate request deletion of their recording, and what is the process?

The last three are the ones vendors answer least comfortably, which is exactly why they are worth asking. Ask us the same questions before you decide.