A frustrated job seeker turned an automated screening call into a conversation between two AI systems, offering a compact picture of what recruiting can become when both employer and applicant automate the interview.
Christopher, identified only by his first name because his search is continuing, told Wired that he had applied for about 700 jobs over six months after work for government contractors dried up. Five applications led to calls from “Riley,” an AI recruiter used by IT firm Everforth Apex Systems.
The first four calls gathered information about work authorization and professional experience, then promised a human recruiter might make contact. None produced a follow-up conversation, email or automated rejection. When Riley approached him a fifth time, Christopher placed the recruiter next to ChatGPT Voice and instructed the chatbot to speak as him.
Two agents complete an interview without progress
The resulting bot-to-bot call lasted ten minutes. Riley and ChatGPT exchanged polite remarks and became caught in a circular discussion of onboarding and background checks. The recruiting agent again said a person would reach out if the qualifications matched. No recruiter did.
Christopher described the exchange as a “slop flywheel”: one synthetic persona supplying synthetic information to another without an apparent outcome. His experiment mixed amusement with dissatisfaction and became a way to reject a process in which repeated automated interviews yielded no feedback.
Everforth Apex Systems did not respond to Wired’s request for comment, so the article does not establish how the company evaluated Christopher’s applications, what happened to the recordings or why no human followed up. It presents his account of the experience rather than an audit of the employer’s hiring system.
Voice screening spreads despite technical limits
AI already appears throughout recruiting, but live voice interviews have been harder to automate. Ophir Samson, head of voice AI at recruiting platform Greenhouse, said agents can struggle with accents and verbal pauses. Even avoiding interruptions—a routine part of human conversation—becomes a difficult engineering problem.
Recruiters nevertheless face high application volumes and are increasingly using voice tools for initial screening. Greenhouse reports that 63% of job seekers have encountered an AI interview. The figure indicates broad exposure, although the Wired story does not say how the survey defined an AI interview or measure the quality of those encounters.
Applicants are also adopting AI assistance. That practice is common enough for recruitment startups such as Ribbon to market tools that identify answers that appear overly scripted, coached or AI-assisted. Automation on one side therefore encourages detection and counter-automation on the other.
Mark Monaghan, vice president of organizational development at call-center company IQor, called bot-to-bot interviewing a logical next stage. He was not enthusiastic about that future, but considered Christopher’s response justified when the employer had sent a bot first.
A perfect candidate produces the same silence
Christopher tested whether his own background, rather than automation, explained the lack of progress. For a sixth Everforth Apex opening, he created a fictional candidate and filled the résumé with qualifications drawn from the job posting.
Riley contacted the fictional applicant immediately. ChatGPT, speaking in that role, spent 23 minutes discussing operational improvement, customer experience and organizational knowledge. It supplied an answer and a personal anecdote for each qualification, then asked its own questions as the interview ended.
The result did not change. Riley repeated that a recruiter might schedule a conversation after reviewing the application, and Christopher heard nothing further.
The episode does not prove that every automated interview is pointless or that the fictional résumé should have advanced. It does expose a failure of closure: a screening system can conduct increasingly polished conversations while giving the applicant no meaningful decision, explanation or human contact. Once candidates can automate their side just as easily, conversational fluency stops being evidence that either participant is learning anything.
