If you've applied to more than a handful of jobs in 2026, you've probably already talked to a machine. Maybe it was a one-way video prompt asking you to record answers with no human on the other end. Maybe it was a live voice call where the "recruiter" turned out to be a conversational AI agent that could ask real follow-up questions. Either way, the AI interviewer isn't an experiment anymore. It's mainstream infrastructure.
According to Greenhouse's own candidate survey, 63% of US job-seekers have now been interviewed by AI - a 13-point jump from just six months earlier. And it's not landing well: 38% of candidates say they've dropped out of a hiring process specifically because it required an AI interview, and another 12% said they'd walk if asked. Here's what's actually happening on the other side of the screen, and how to not be part of that 38%.
Three different things people call "AI interview"
The category is broader than it looks, and the differences matter for how you prepare.
1. One-way async video. You get a prompt, you record an answer alone with your webcam, no one is on the call. Platforms like HireVue and Spark Hire dominate this format. HireVue alone processed nearly 20 million assessments and interviews in a single quarter of 2024. This is asynchronous - you can often retake, and there's no real-time follow-up.
2. Live conversational AI interviewers. A genuinely new category as of 2025-2026: an AI agent conducts a real-time, two-way voice or video interview, asks follow-up questions based on what you say, and produces a scored writeup for the hiring team. Startups like Ribbon (over 1 million interviews run, founded 2023) and Alex AI (formerly Apriora) are the visible players here; Forbes profiled Apriora's "Alex" as early as mid-2024, and by 2025 the category had enough traction that Bloomberg covered it as a mainstream hiring tool, not a novelty. Growth has been fast on the vendor side too - Ribbon reportedly signed close to 400 customers in its first eight months of selling the product, which tells you how much appetite there is on the employer side to offload first-round screening entirely.
3. AI-assisted human interviews. A human recruiter still runs the call, but AI tools transcribe, summarize, and flag competency signals in real time. This is the least visible category to candidates and the fastest-growing one inside HR teams broadly. It's a separate trend from the shift happening in later-stage technical rounds, where the pressure is pushing the opposite direction - toward more in-person, human-only formats. Our breakdown of where the coding interview stands in 2026 covers that side of the pipeline.
Across all three, adoption jumped fast: AI use in HR tasks climbed from 26% of organizations in 2024 to 43% in 2025, and 96% of hiring professionals now use AI in at least some part of recruiting, per SHRM's 2026 State of AI in HR research.
How the scoring actually works
The mental model most candidates have - a robot judging your smile and eye contact - is mostly outdated, at least for the biggest player. HireVue quietly discontinued facial analysis scoring in March 2020. CEO Kevin Parker later explained the visual signal simply stopped adding predictive value once language-based analysis improved: "the visual components contributed less to the assessment to the point where there was so much public concern to everything related to AI that it wasn't worth the concern it was causing people." Chief Data Scientist Lindsey Zuloaga added that candidates were "never evaluated based on sensitive or personal visual attributes."
What's actually being scored today, for one-way video platforms, breaks down into three layers:
- Verbal content - word choice, whether you name specific competencies and concrete examples instead of generic claims.
- Vocal delivery - pace, confidence, and hesitation patterns in your speech, not your face.
- Structural response quality - whether your answer has a clear shape: situation, action, result. This is the layer most in your control, and it's why the STAR method isn't just generic interview advice anymore - for an AI grader, it's close to a literal formatting spec.
Live conversational AI interviewers work differently: because they can ask a real-time follow-up, a memorized script falls apart the moment the agent probes deeper. The evaluation is closer to a real interview, just conducted by an agent instead of a person.
You increasingly have a legal right to know
This isn't just a courtesy question anymore in some jurisdictions - it's a growing legal requirement.
New York City's Local Law 144, in force since July 2023, requires any employer using an "automated employment decision tool" to substantially assist or replace hiring decisions for NYC-based roles to: commission an independent bias audit annually, publicly post a summary of the results, and give candidates at least 10 business days' notice before the tool is used to evaluate them, along with the option to request an alternative process. Enforcement carries civil penalties of $500-1,500 per day per violation, and each missed notice counts separately.
Illinois has a parallel law requiring employers to notify candidates and employees when AI is used to make employment decisions about them. The broader trend across new state legislation is the same: disclosure of AI involvement, and a legal record of what the tool evaluated you on.
Despite this, transparency in practice still lags: research summarized in HR Dive and Fortune's 2026 coverage of AI interviewing found the large majority of candidates weren't clearly told in advance that AI would be involved, and a meaningful share only realized it once the interview had already started. If you're in a jurisdiction with a disclosure law, it's worth directly asking the recruiter: "will this interview be scored by an automated tool, and can I get the required notice?" You're often legally entitled to ask.
How to actually pass one
Treat structure as the rubric, literally. For one-way video and most live AI interviewers, a tight STAR-shaped answer scores measurably better than a rambling one, because structure is one of the few signals the system can grade with confidence. Our behavioral question framework works directly here - build 5-6 universal stories in advance and adapt them, rather than improvising live.
Say the keywords, don't imply them. Human interviewers infer competencies from context. AI graders weight explicit language more heavily. If the question is about ownership, use the word "owned," not a story that vaguely gestures at it.
Fix your setup before the interview, not during it. A broken audio track or laggy video produces an incomplete transcript, and incomplete data scores lower by default regardless of what you actually said. Test your mic, lighting, and connection the day before.
For live conversational AI interviewers, prepare like it's a real interview, because it functionally is one. These agents ask real follow-ups. A canned answer that would survive a human's polite nod will get probed by an agent that's specifically built to dig one layer deeper. Practice out loud, not just in your head - our interview prep checklist has the full run-through routine.
Ask about the format up front. Whether you're facing an async HireVue-style prompt or a live agent changes your prep completely. If a recruiter doesn't volunteer which one it is, ask - in NYC and Illinois you may be entitled to know anyway.
Pull real questions for your specific role. Generic STAR practice only gets you partway. Pair it with the actual prompts asked for your target role - our role-specific interview question guides are a good place to build your answer bank before you're staring at a recording prompt with a 90-second timer.
The bottom line
The AI interviewer isn't going away, and pretending it's still 2022 - where the biggest risk is bad lighting - will cost you. The systems doing the actual scoring today grade on content and structure, not your face, and they're increasingly required by law to tell you they're involved. Prepare answers with real structure, confirm the format before you show up, and use your growing legal right to ask what's evaluating you.
None of this means the AI round is the whole process. Most pipelines still route a shortlist to a human for a final decision, and a well-structured AI screen is usually just the first filter, not the last word. Treat it with the same seriousness as a human phone screen, because functionally, that's exactly what it replaced.
Sources
- 63% of Job Seekers Have Faced an AI Interview. Most Haven't Had a Good One Yet - Greenhouse
- Nearly 4 in 10 job candidates have bailed on a hiring round because it required an AI interview - Fortune
- HireVue Discontinues Facial Analysis Screening - SHRM
- The State of AI in HR 2026 Report - SHRM
- Automated Employment Decision Tools (AEDT) - NYC Department of Consumer and Worker Protection
- Your Next Job Interview May Be With 'Alex,' The AI Interviewer - Forbes
- Job interviews enter a strange new world with AI that talks back - Bloomberg (via Detroit News)
