AI Hiring Interviews: What Candidates Should Know Now12 min read2,469 words

AI Hiring Interviews: What Candidates Should Know Now

AI is changing how companies screen candidates—and how they judge applications. Learn what the hiring data shows and how to respond as a candidate or employer.

AI hiring interviewsAI recruitmentAI resume screeningethical AI hiringcandidate experience
AI Hiring Interviews: What Candidates Should Know Now

AI Hiring Interviews Are Growing. Here's What That Means for You

A candidate can run into a maddening contradiction: employers may distrust an application that seems AI-generated, but still ask that same candidate to interview with an AI system. That tension matters because hiring pipelines decide who gets heard - and early screening can shape a candidate's chances before a recruiter ever reads or speaks with them.

Greenhouse survey data gives a snapshot of the problem, with all the usual caveats. Nine in ten surveyed US employers said they believed they had turned away real candidates because their applications seemed fake or AI-generated. At the same time, half of employers said they use AI interviews to reach more applicants. These are employer perceptions and reported practices, not independently measured hiring outcomes.

So why do these things show up together, what does the available data actually tell us, and how can candidates and employers make AI interviews more useful and fair? The practical question isn't whether AI belongs in hiring. It's how to use it without swapping polished writing for real ability - or automated screening for sound judgment.

Why AI suspicion and AI interviews coexist

At first glance, the numbers look hypocritical. Employers say AI-generated applications make it harder to identify genuine candidates, yet they're also using AI to conduct more interviews. The contradiction is real - but it comes from different pressures inside the hiring process.

Generative AI makes it easier to produce applications that are fluent, tailored and consistent in tone. That can help a qualified candidate explain their experience. It can also make it harder for a recruiter to tell whether an application reflects the candidate's real skills. One response is to treat unusually polished - or overly generic - applications with suspicion, even when the candidate wrote them without AI.

Then there's the other pressure: volume. When a role attracts hundreds of applications, a human team often doesn't have enough time to speak with everyone who meets the basic requirements. An AI interview can act as a structured first step for more applicants, without requiring a recruiter to schedule and conduct every screening call.

More access does not automatically mean better screening

Greenhouse reported that half of employers now use AI interviews to hear from more candidates. Among employers that use the technology, 40% said they had found strong candidates they would otherwise have missed. That sounds like a real benefit: a wider first-stage conversation can give candidates another way to show relevant experience beyond a résumé.

But a bigger pool isn't the same thing as a fairer or more accurate process. If the system asks the wrong questions, misreads an answer or makes the interview hard to complete, it can scale a bad experience just as effectively as a good one. And if recruiters treat an automated summary like a final verdict, candidates can get screened out without meaningful human review.

The job market changes the meaning of "choice"

An applicant may accept an AI interview because it's the only route to the next stage - not because they actually prefer it. In an earlier Greenhouse report, only 7% of US applicants said they would turn down an AI interview if given the choice. But that doesn't mean candidates were broadly comfortable with the idea. The same report found that 81% of job seekers were rejected in the previous year without speaking to anyone at the company, whether human or AI.

When the alternative is no conversation at all, a short automated interview can feel like the only workable option. That still doesn't remove the employer's responsibility to explain what's happening and provide a reasonable way to raise concerns. Access only helps when the process stays understandable, relevant and open to human review.

What the hiring data does and does not show

The Greenhouse findings come from a survey of 1,347 hiring managers and recruiters in the US, UK, Germany, Australia and Ireland, conducted from late September to mid-October. The results help explain what respondents say they believe and do. They don't amount to a controlled study of how often AI is actually responsible for rejecting candidates.

Greenhouse's head of Voice AI, Ophir Samson, told ZDNET that the finding about rejected applications is self-reported. In other words, employers said they believed they had turned away real candidates because an application seemed fake or AI-generated. The survey doesn't show whether employers used AI-detection tools, what criteria they used, or whether the rejected candidates were truly unqualified.

Employer perceptions are not measured outcomes

That distinction matters. A survey response can show a real concern, but it can't prove the cause of a hiring decision. A recruiter might suspect an application was generated by AI even when the candidate wrote it themselves. Or an application might have been rejected for a different reason, with AI suspicion remembered afterward as part of the explanation.

The same limitation applies to claims that AI interviews improve the candidate experience. Half of employers using voice AI said the early-stage experience had improved, but the survey doesn't define what "improved" means or whether candidates would rate it the same way.

Greenhouse also has a commercial interest in AI hiring tools, including a voice interview product. That doesn't make the survey findings useless, but it does mean vendor-reported results should be treated as one input - not a neutral verdict. Better decisions require more than a headline percentage. Employers should measure their own process, and candidates should understand how their responses are handled.

Human oversight is a design choice, not a guarantee

Samson said Greenhouse's Voice AI doesn't automatically advance, reject or rank candidates, and that recruiters review every output. He also described an example where a recruiter used the tool to revisit 200 candidates who had already been rejected. The system suggested 19 for further consideration, and all five eventual finalists came from that group. The employer ultimately hired one of them.

That example points to a plausible use case: AI can help surface candidates who might be missed in an initial pass. But it doesn't tell us how often the system misses good candidates, whether similar results would happen at other companies, or how recruiters decided which recommendations to act on. A "human in the loop" only helps if the reviewer has the right information, time and authority to challenge what the tool produces.

For employers, the real standard isn't just "a human reviews it." Ask whether the reviewer can see the original response, understands what the system can and can't do, and can document why someone advances. For candidates, look for clear explanations of what's automated and who makes the final decision.

How candidates can approach an AI interview

An AI interview is still an interview. You still need to make your experience clear, connect it to the role, and provide specific evidence. The difference is that you might not get the conversational cues - or the follow-up questions - you'd expect from a human interviewer. That changes how you should prepare.

Skip trying to guess or "game" the scoring. The system's design may be opaque, and answering based on imagined rules can make your responses less useful. Instead, prepare concise examples that show what you did, what constraints you worked under, and what changed because of your actions.

Prepare examples that stand on their own

Pick a few relevant examples from your work, and build each one around four parts: the situation, your responsibility, the actions you took and the outcome. A concrete story about solving a production issue or improving a customer workflow gives a reviewer more to evaluate than a chain of vague claims like "strong communicator" or "results-driven."

Because an automated interviewer may not ask a natural follow-up, you may need to supply context a human would normally pull out of you. Explain what your contribution was, what trade-offs you considered and how you measured success. Keep it focused - detail helps when it supports the point, but a long answer can bury the evidence.

A candidate described by Samson was initially treated as a "maybe" on paper. In the interview, he was able to mention skills he hadn't included on his résumé. That's not proof that AI interviews reliably find overlooked talent. But it does show how a conversation can add information a résumé doesn't carry.

Check the process before you start

Before recording an answer, try to find out what you can about the format. Reasonable questions include:

  • Will a human review my responses?
  • How will my answers be evaluated?
  • Is the interview timed, and can I retake or correct an answer?
  • Is there an alternative format or a way to request an accommodation?
  • How will my recording or transcript be used and retained?

Greenhouse reported that nearly half of job seekers wanted to know whether a human would review their responses, and 40% wanted to know how they would be evaluated. Those aren't unreasonable expectations. Clear answers help candidates decide what they're agreeing to - and help employers earn trust in the process.

If you need an accommodation, run into a technical problem or feel uncomfortable with the format, contact the recruiter rather than quietly dropping out. Whether an alternative is available will depend on the employer and jurisdiction, but asking creates a record and gives the company a chance to respond.

How employers can use AI without outsourcing judgment

AI interviews can help when the bottleneck is simply time: too few people to handle basic conversations. They're a poor substitute for defining the role, choosing relevant criteria and making accountable decisions. The goal should be to give recruiters more capacity to assess candidates - not to turn a model's output into an unchecked ranking.

Greenhouse reported that 86% of surveyed employers changed their hiring processes in response to AI's impact, and 57% of that group added another human interview. That may increase scrutiny, but more stages also means more time and work for candidates and recruiters. More "touchpoints" aren't automatically better if each stage repeats the same questions without adding useful evidence.

Design the decision before selecting the tool

Start with the hiring decision the interview is meant to support. Is it checking minimum qualifications, collecting examples a recruiter can review, or assessing a skill that can't be established from an application? If you can't answer that clearly, automating the interview is likely to add process without improving the decision.

Then set boundaries:

  • Keep criteria job-related. Tie questions to the work, and use a consistent rubric to compare responses.
  • Tell candidates what's happening. Explain that AI is involved, what it does and whether a person reviews the output.
  • Use the original evidence. Review the candidate's actual response, not only a generated summary or score.
  • Provide a route for exceptions. Candidates need a way to report technical problems, request accommodations or ask for another format.
  • Audit outcomes. Track who advances and who drops out, including whether the process creates patterns the team can't justify.

These controls take effort, but they also make it easier to catch a bad question, a misleading summary or a workflow that filters people for reasons unrelated to the job. It's generally cheaper to fix early than after the tool becomes embedded in a high-volume process.

Keep a named person accountable

In the Greenhouse survey, only 7% of employers said they would hand an AI agent the job of deciding which candidates should advance. That suggests employers still see advancement as a human responsibility. The key is making that responsibility real: someone needs the authority and context to disagree with the system.

A workable workflow should log what the tool contributed, what the recruiter reviewed and why the decision was made. It should also create feedback for improvement. If recruiters repeatedly override a summary, investigate whether the system is failing, the criteria are unclear or additional training is needed. If strong candidates consistently get missed early, revisit the screening process.

Greenhouse also said that among employers that saved time with AI, 85% put that time back into candidates - often by spending longer with shortlisted applicants or interviewing more candidates earlier. That's the right way to measure efficiency: not just whether the team processes more applications, but whether saved time improves the quality and fairness of human evaluation.

Frequently Asked Questions

The survey reports that many employers believe they have rejected real candidates for this reason, but it does not show how those decisions were made or whether an AI detector was involved. Candidates should focus on specific, verifiable evidence of their experience rather than trying to make an application appear less polished.
That depends on the employer and the tool. Greenhouse said its Voice AI does not automatically advance, reject or rank candidates, and recruiters review its outputs. Other systems and workflows may differ, so candidates should ask what is automated and who makes the final decision.
There is no universal answer. Consider whether the employer explains the process, whether a human reviews your responses and whether the format works for you. If you have a concern, ask about alternatives or accommodations before deciding. The available options depend on the employer and the circumstances.
Not by default. An interview can give candidates a chance to explain skills missing from a résumé, but an automated process can also misread answers or create new barriers. Fairness depends on job-related questions, accessible design, clear evaluation criteria and meaningful human review.

Related articles