An AI-conducted interview is not a video presentation tip or a recorded selfie clip a recruiter watches later. It's a screening step where a chatbot or a digital avatar, not a person, actually asks the questions, sometimes including follow-ups, and the system produces a score before a human ever joins the process. It's become common enough in 2026 that candidates applying to larger employers should expect to encounter one at some point in the pipeline, particularly at the earliest screening stage where a human recruiter's time is most limited.
What actually happens in an AI-conducted interview
The format varies by vendor: a text-based chat, a phone call, or a video chat with a photorealistic on-screen avatar. Whichever form it takes, the workflow is structured and consistent: you get a set of role-specific questions, often the same questions every candidate for that role sees, and you respond by typing, speaking, or talking to the camera. Because there's no human on the other end, scheduling is usually available around the clock, and some systems ask adaptive follow-up questions based on what you just said rather than sticking to a fixed script. This also means every candidate for a given role tends to see the same base questions and gets measured against the same standard, which is part of the appeal for employers running high volumes of applicants through an early screening stage.
What the AI is evaluating
The system compares what you say against a rubric built for the role you applied for, often trained on patterns from candidates who previously performed well in that position. It's looking for concrete examples, specific numbers or outcomes, and relevant technical detail, not generic phrases that could apply to any job. A vague answer that would sound fine in casual conversation tends to score poorly here, because there's nothing specific for the model to match against the rubric. Structure matters too: an answer built around a clear situation, action, and result tends to translate cleanly into whatever summary a recruiter eventually reads, while a meandering answer is harder for both the model and the human reviewer to credit fairly.
Where a human still makes the final call
Despite the automated framing, most AI-conducted interviews are not built to reject or hire on their own. Regulations like New York City's Local Law 144, which requires bias audits for automated employment decision tools, and the EU AI Act, push vendors toward a decision-support model: the AI produces a score and a summary, and a recruiter or hiring manager still reviews it before deciding who moves forward. That human step is worth knowing about, since it means a strong, specific answer that a machine scores well is also the kind of answer a person reviewing the summary will find convincing. It's also why disclosure is becoming more common: many employers now say upfront, in the invitation email or on the screening page itself, that part of the process is AI-conducted, rather than leaving candidates to guess whether they're talking to a person.