What Candidate AI Tools Reveal About Your Screening Call

What should a hiring team do when candidates start bringing AI to the screening call? I mean software that listens to the recruiter's questions and produces answers in real time, which the candidate reads off a screen or hears through an earpiece. These tools have spread quickly, and if you run a lot of phone screens you may already have talked to candidates using them without knowing.

The first reaction most talent leaders have is to ask how to detect it. I understand why, but I think it's the wrong first question. Before worrying about whether an AI got through your screen, it's worth asking whether your screen was telling you much to begin with.

Think about what a typical first-round phone screen is. Fifteen or twenty minutes, built on questions that have been posted on Glassdoor and Reddit for years: tell me about yourself, what's your greatest weakness, why are you interested in this role. Those questions reward preparation, and every serious candidate already prepares for them. So an AI that answers them fluently hasn't beaten anything. A well-rehearsed person would have given much the same answers, and you'd have learned about as little.

Seen that way, the AI tools didn't create a weakness. A process that filters people by how well they interview was always open to gaming, and coaching and rehearsed answers were gaming it long before any software did. AI just made it cheap and available to everyone.

How far has it gone? There have been reports of candidates getting through multi-stage interview processes at large employers with a tool feeding them answers, and apps built for exactly this are openly available, some claiming hundreds of thousands of downloads. None of it is sophisticated technology. It doesn't need to be.

The more useful question is which kinds of screening these tools beat easily, and I'd expect the answer to be the fixed ones. An application questionnaire, an async video with preset questions, a chatbot that walks every applicant through the same qualification flow: each of these gives the tool a known target. When the questions are fixed in advance, you can prepare a perfect answer to each one, and so can software.

A live spoken conversation is harder, especially one where each question depends on what the candidate just said. Suppose a candidate says they have four years of experience in some area. A good interviewer's next question is about that claim specifically. If they describe a project, the follow-up asks about a decision they made halfway through it. Nobody wrote that question before the call started. To handle it, an AI would have to follow the whole exchange, stay consistent with everything said so far, and answer out loud in real time, in a conversation whose shape it can't predict. That's a much harder problem than producing a polished paragraph about your greatest weakness.

The same property is what makes the conversation worth having in the first place. Answering out loud, on the spot, to questions that follow from your own words is a lot closer to what most jobs actually ask of people. So the formats that are easiest to game also tend to tell you the least, and the ones hardest to game tend to tell you the most about how someone thinks. I don't think that's a coincidence. A format that rewards real thinking is hard to fake for the same reason it's informative.

That's the kind of screen we built Asendia to run. We make software that phones applicants soon after they apply and interviews them, with each question following from the last answer, as in the example above. Catching cheaters wasn't the goal. The goal was to learn as much as possible from a first conversation, and being hard to fake came along with that. The same problem also shows up one step earlier, in the AI-generated applications flooding your ATS.

So I wouldn't treat candidate AI agents as a crisis. They're more like a test result, and what they show is that the screens they beat were already measuring the wrong thing. Teams that set out to catch AI use are signing up for a contest they'll probably lose, since the tools are cheap and keep getting better. Teams that ask what the tools reveal about their screening, and change the first conversation because of it, end up with a process that's harder to game and better at predicting who will do well.

Ready to transform your hiring strategy? Schedule a Demo with our founders today!