38% of Job Seekers Walk Away From AI Interviews — Here's How to Pass Them

The Greenhouse 2026 Candidate AI Interview Report surveyed nearly 3,000 active job seekers and found that 63% have now been interviewed by an AI — up 13 percentage points from just six months earlier. The same report found that 38% of candidates have already walked out of a hiring process because it involved an AI interview, and another 12% say they would.

But walking away isn't always an option. If you're deep in a process at a company you actually want to work for, opting out of the AI round often means opting out of the job. So the more useful question isn't "should I do this?" — it's "how do I pass it?"

Here's what's actually happening, what these systems look at, and what you can do about it.

What "AI Interview" Actually Means

The term covers at least three distinct formats, and the strategies differ by type.

One-way video interviews are the most common. You're sent a link, shown a question, given 30–60 seconds to prepare, and then record a video response. No human sees it in real time — an AI scores it first. Platforms like HireVue, Spark Hire, and Modern Hire (now merged into HireVue) use these at scale. If you score above a threshold, your video gets routed to a human recruiter.

Chatbot-driven text interviews replace the initial phone screen with an async conversation. You type answers to structured questions about your background and experience. These are typically scored on keyword presence and answer completeness, then summarized for a recruiter.

AI-scored live video is the newest format — you're on a video call with what feels like a real interview, but the "interviewer" is a conversational AI that responds to your answers and probes deeper. Some companies use human interviewers while AI scores the conversation in the background.

Knowing which format you're walking into changes how you prepare. Ask the recruiter: "What format is the interview?" and "Will I be speaking with a person, or is this an automated process?" Companies that are doing this ethically will tell you.

Why the Resistance Is Justified — and Why It Doesn't Change the Math

The backlash is real. The Greenhouse report found that 70% of candidates were never clearly told upfront that AI would be evaluating them, and one in five only found out once the interview had already started. Fifty-one percent of those who completed an AI interview never heard back at all — no rejection, no update. Half of all candidates in 2026 have been rejected at least once without a single word from a human.

That experience — preparing, performing, and then disappearing into a void — is genuinely demoralizing. Fifty-seven percent of job seekers now believe disclosure should be a legal requirement, and some states are beginning to consider it.

The structural unfairness is real. But the math of job searching is also real: if the company you want to join uses this process, the choice is to play or withdraw. Most candidates, most of the time, choose to play.

What the Algorithms Actually Evaluate

The scoring criteria vary by platform, but the core inputs are consistent across most enterprise-grade AI interviewing systems.

Content and structure are the heaviest factor in text-based scoring. The AI is looking for relevance to the question, specific examples rather than generalities, and complete answers. "Tell me about yourself" scored by an AI wants to see a career arc, a current role, and a future direction — not a memoir.

Word choice and vocabulary matters more than most candidates expect. Vague language ("I kind of helped with that project," "we sort of turned things around") scores lower than specific language ("I led the rollout," "revenue grew 18%"). Filler words — "um," "like," "you know," "basically" — are logged and factored in.

Vocal pacing and confidence are scored in one-way video. Speaking too fast, running sentences together, or audible hesitations after questions all affect the signal. The system isn't reading your nervousness the way a human would — it's logging pattern deviations from a trained baseline.

Facial cues are used by some platforms (notably HireVue's earlier versions, though their current use is more limited following regulatory scrutiny). Eye contact with the camera, stable positioning in the frame, and natural expression are all inputs. This is contested territory, and some jurisdictions have restricted AI facial analysis in hiring.

Response completeness for text-based formats means giving an answer that's long enough to demonstrate substance but not so long that it buries the point. Most platforms have an implicit length model. Too short (one or two sentences) signals low engagement. Too long without structure signals poor communication.

Seven Things That Actually Move the Score

1. Know the format before you start. Ask upfront. If the company won't tell you, that's information too.

2. Lead with the outcome. AI scoring systems, like human recruiters pressed for time, give more weight to what they read or hear first. "I reduced churn by 22% over six months, and here's how" is a better opening than a three-paragraph setup before the result.

3. Use explicit structure. STAR (Situation, Task, Action, Result) works well precisely because it produces a clearly segmented answer. Labeling the structure out loud — "The situation was…, and the result was…" — also helps text-scoring systems parse it correctly.

4. Cut the filler. Record yourself answering three practice questions and count your filler words. It's the fastest lever. One session of deliberate practice reduces them significantly.

5. Look at the camera, not your image. In one-way video, there's no other person to make eye contact with — so candidates default to watching themselves, which reads as poor eye contact. Put a small sticker next to the camera dot and keep your gaze there.

6. Stay under two minutes per answer. Most one-way platforms have a time limit. Getting cut off mid-answer is a hard fail. Practice delivering your three or four strongest examples within 90 seconds.

7. Eliminate technical distractions in advance. AI interview scores are affected by audio quality, background noise, and lighting — the system is calibrating based on audio clarity. Test your mic, front-light your face, and use a neutral background. These take ten minutes and eliminate entirely preventable score drags.

Prepare Like the Algorithm Is Watching — Because It Is

The irony of AI interviews is that the best way to prepare for them is to practice with AI. Before a high-stakes AI interview, load your resume, the job description, and a list of the company's values or recent news into a prep session. Work through common questions out loud. Review your word choice, your answer structure, and your pacing.

Meeting Copilot's interview assistant lets you run timed practice sessions where you can hear yourself back, spot filler patterns, and refine your strongest examples before they count. And for processes that include a live human interview after the AI screen, having context-aware suggestions in real time closes the gap between your rehearsed best and your pressure-state performance.

You Don't Have to Accept Every AI Interview on Its Terms

You have more leverage than the current landscape makes it feel like. You can ask if there's an alternative. Some companies, especially when hiring senior roles or in competitive markets, will offer a phone screen instead. You can ask whether the AI scoring will be the sole factor in routing you to the next round, or whether a recruiter reviews all responses. These are reasonable questions that a good hiring team should be able to answer.

You can also simply withdraw. Fifty-seven percent of candidates believe disclosure should be required; some argue that refusing to participate in undisclosed AI evaluation is a principled stance worth taking, and that companies doing this poorly will eventually feel the candidate experience consequence in their offer acceptance rates.

The Greenhouse data suggests that's already starting to happen. Fifty percent of candidates have been rejected without hearing from a human being. Employers that can't close that feedback loop will keep losing candidates to companies that can.

The Bottom Line

AI interviews are mainstream and getting more common. Most candidates in 2026 will encounter at least one. The systems aren't perfect, the transparency isn't there yet, and the candidate experience leaves a lot to be desired — but the skills they're actually testing for (clear communication, structured thinking, specific examples, outcome-first answers) are real skills worth having.

The best preparation for an AI interview is the same as the best preparation for any interview: know your material cold, structure your answers clearly, and practice until the delivery feels effortless.

The difference now is that the evaluator doesn't give you the benefit of the doubt. Build in the work the system can actually see.