Art Hebbeler did not need a second take. After years in media and communications, he logged into what he expected to be a first round conversation and found an AI interviewer waiting instead. No hello, no small talk, just prompts on a screen and a countdown timer. He described the experience as cold and stilted, the kind of interaction that tells you how a company values people before you ever meet one.
Tammy Wright had a similar moment from the other side of the market. A seasoned recruiter herself, she was asked to interview with an AI system for a role she wanted. She declined. Not because she fears automation, but because she did not want to be assessed by a system that could not answer her questions, read context, or explain how it would judge her.
They are not outliers. In a survey of 1,200 U.S. job seekers, nearly 4 in 10 have withdrawn from a hiring process because of an AI interview, and nearly two thirds said they have already had one. A separate Greenhouse survey of 2,950 active job seekers found 63 percent have been interviewed by AI, up 13 percent in the previous six months. Candidates are not just complaining about AI interviews. They are walking away from them, and in some cases blacklisting the employers who use them.
The efficiency myth that sold the AI interview
Hiring teams did not adopt AI interviews to be impersonal. They adopted them to cope with volume. The number of U.S. applicants per open role has doubled since spring 2022, and recruiters are under pressure to screen faster without adding headcount. In a January LinkedIn survey of more than 6,500 HR professionals, two thirds of recruiters said they plan to increase their use of AI for pre-screening interviews this year. On paper, the logic is clean. Let software handle the first conversation so humans can focus on the final one.
At the market level, that efficiency has not shown up. Artificial intelligence has not consistently improved hiring speed despite 91 percent of HR leaders saying they actively use AI in recruitment, according to ManpowerGroup research released September 8, 2026 covering nearly 40,000 employers across 42 countries including more than 6,000 U.S. employers. Only a third of employers reported improved time to hire compared to 2025, while 42 percent reported no change and 25 percent reported the process actually slowed down. A lengthy recruitment process was cited as a key obstacle to hiring, not a problem solved.
That gap between perception and outcome helps explain why teams keep adding more automation even as candidates pull back. When 70 percent of hiring managers say AI helps them make faster and better hiring decisions with fewer recruiter resources and one in two recruiters say AI has improved hiring overall, the internal feedback loop feels positive. The external feedback loop tells a different story. Speed that is measured inside the ATS does not count if it is lost to candidate dropout outside it.
| What hiring teams expect from AI interviews | What the market data shows |
|---|---|
| Faster screening and shorter time to hire | Only 33% reported faster hiring, 25% reported slower hiring |
| Better decisions with fewer resources | 70% of managers believe it, 8% of candidates agree it is fairer |
| Capacity to handle double the applicants per role | 42% saw no change in speed despite 91% adoption |
Why the pressure to add more AI keeps rising
Volume is real, and it is not going away. When every open role draws twice as many applicants as it did three years ago, any tool that promises to watch every video or read every transcript looks like relief. The problem is that pre-screening interviews are not just a volume problem. They are a trust problem. Candidates decide in the first ten minutes whether an employer is worth their time, and an autonomous interviewer often fails that test even when it saves a recruiter an hour.
The validity problem behind the backlash
Candidates describe AI interviews as dehumanizing, but the deeper issue is validity. The systems are good at scoring what is easy to encode and weak at assessing what actually predicts performance.
New City & Guilds data makes the mismatch explicit. 69 percent of employers say AI powered screening tools prioritize technical skills over human skills like communication, problem solving and time management that they say predict job performance, while 43 percent already use AI to screen or filter candidates. At the same time, 74 percent of employers now rate employability skills as more important than technical expertise, yet 64 percent say candidates with strong employability skills have become harder to find. In other words, teams are deploying tools that over weight keywords and under weight judgment, then wondering why the shortlist feels thin.
The same pattern shows up in interviews. Systems that score eye contact, speech rate, or keyword density can create an illusion of objectivity while missing the signal that matters. A candidate who pauses to think, who asks a clarifying question, who adapts an answer to the interviewer, is demonstrating the very skills employers say they want. An AI interviewer that cannot respond in kind will often penalize that behavior or ignore it entirely.
Trust is not a soft metric in hiring. When candidates do not trust the interview, they do not finish the process.
Governance has not kept pace either. More than a quarter, 28 percent, of employers using automated screening do not monitor whether outcomes differ by ethnicity or other protected characteristics, and of those who do check, 82 percent found differences and 36 percent described them as meaningful. That is not a hypothetical bias risk. It is a measured outcome gap with limited oversight, and it helps explain why only 8 percent of candidates believe AI makes hiring fairer while hiring managers remain confident it does.
What candidates actually object to
The backlash is often framed as candidates being anti AI. The complaints are more specific, and more fixable, than that.
- No feedback loop: Candidates answer into a void. They cannot tell if an answer landed, they cannot course correct, and they receive no feedback on why they were rejected. The experience feels like an assessment, not a conversation.
- No ability to ask questions: A first round interview is also a candidate's interview of the employer. When the interviewer is software, there is no one to ask about the team, the work, or what success looks like. That asymmetry signals that the employer is screening, not selecting.
- Accessibility and dialect risk: Candidates worry, with reason, that systems trained on narrow speech patterns will misread accents, dialects, or disabilities. Even when vendors claim otherwise, the lack of transparency makes it hard to trust.
- Proxy for culture: Candidates read process as culture. If an employer shortcuts the human touch at the first human moment, they assume it will do so later on onboarding, management, and development. Walking away is a rational hedge.
- Black box scoring: When scoring criteria are not disclosed, candidates cannot prepare meaningfully or contest errors. The perception, fair or not, is that the system is looking for reasons to filter out, not reasons to select in.
These objections compound. A candidate who is unsure how they are being judged, cannot ask questions, and suspects the system is biased has little incentive to continue, especially in a market where they can apply elsewhere in minutes. The 4 in 10 withdrawal rate is not a protest vote. It is a response to a process that feels low signal and high risk.
A better playbook for interview stage AI
The alternative is not to ban automation. It is to move it from the center of the interview to the edges, where it can support better human decisions instead of replacing them.
1. Disclose where AI is used and why
Tell candidates before the interview what is automated, what is not, and how their data will be used. Is AI generating questions, transcribing, summarizing, or scoring. Is a human making the decision. Candidates tolerate automation far better when the rules are clear. Transparency also forces internal discipline about what the tool is actually for.
2. Keep a human as the decision maker
Use AI to prepare and capture evidence, not to autonomously score or reject. That means structured question generation tied to the role, real time transcription and evidence capture, and post interview summaries that a human reviews. The hiring manager still decides, with a traceable record of why.
3. Design for employability skills, not just keywords
If communication, problem solving, and time management predict performance, the interview must elicit them. Build structured interviews with behavioral prompts, follow ups, and work samples that let candidates show judgment. Use AI to ensure every candidate gets the same core questions and that interviewers probe consistently, rather than to count buzzwords.
4. Audit selection rates by stage and group
If you use AI anywhere in screening or interviewing, monitor outcomes by ethnicity, gender, disability, and other protected characteristics at each stage. The finding that 82 percent of employers who check find differences should be a prompt to check earlier and more often, not to look away. An audit that finds a meaningful gap is not a failure. It is the system working as intended, giving you a chance to fix the stage before it becomes a pattern.
5. Measure trust as a funnel metric
Track completion rates, withdrawal reasons, and candidate Net Promoter Score for AI involved stages the same way you track pass through rates. A stage that saves 20 minutes per candidate but loses 15 percent of qualified candidates is not efficient. It is expensive.
Speed without trust is not speed. Candidates who drop out never become hires.
The takeaway for TA leaders
We are in a low hire, low fire market where every qualified candidate matters more. ManpowerGroup found that hiring speed has not improved for most employers even as AI adoption hit 91 percent. City & Guilds found that the skills employers value most are the ones AI screens are most likely to miss. And candidates are telling us directly that they will not sit for a process they do not trust.
That combination reframes the AI interview from a cost saver to a brand and selection decision. The cost is not just the license. It is the candidates who never make it to a human, the signal you lose when you score the wrong things, and the trust you erode when you automate the most human part of hiring.
The competitive advantage now is not a faster filter. It is a more credible process. Use AI to make interviews more structured, more consistent, and better documented, but keep people at the center of the conversation. Disclose the role of AI, design for the skills that predict performance, and audit what happens at each stage. In a market where talent is cautious and volume is high, a transparent, human led interview is not just the right thing to do. It is the fastest way to actually hire.
