What Happens When Candidates Know They’re Talking to AI?
By Aadisri Walia ·
A candidate finished her AI screening interview last month and left a comment in the feedback box: “I actually said more than I would have with a real person. No one was judging my pauses.”
That line stuck with me.
We spend a lot of time talking about AI screening from the recruiter’s side—efficiency, consistency, time saved. But there’s a question that comes up in almost every conversation with TA leaders considering this stuff: What happens on the candidate’s end? Do they clam up? Game the system? Resent the process?
The disclosure question
First, the practical matter: should you tell candidates they’re talking to AI?
Yes. Full stop.
This isn’t just an ethical position—though it is that. It’s also a legal one in several jurisdictions. Illinois, Maryland, and New York City all have laws requiring disclosure when AI is used in hiring decisions. The EU’s AI Act is pushing similar requirements. Even where it’s not legally mandated yet, the direction is clear.
But beyond compliance, there’s a trust argument. Candidates talk. They post on Reddit, Glassdoor, Blind. If someone discovers mid-process—or after—that they were screened by AI without knowing, you’ve created a reputation problem that outlasts any efficiency gain.
Disclose clearly, disclose early, and don’t bury it in paragraph nine of your privacy policy.
The assumption that doesn’t hold
Here’s what most people expect: tell candidates they’re talking to AI, and they’ll either check out emotionally or try to game the system. The interaction becomes artificial. You lose the signal you were hoping to capture.
The more common reaction is something closer to relief. Candidates report feeling less nervous, not more. Less worried about awkward silences. Less anxious about saying “um” too many times or needing a moment to think.
There’s no human on the other end forming a snap judgment about their outfit, their accent, their Zoom background. Just a structured set of questions and time to answer them.
One candidate put it this way: “I didn’t feel like I had to perform confidence. I could just… answer.”
Why authenticity goes up, not down
This seems counterintuitive until you think about what makes traditional interviews stressful.
It’s not the questions. It’s the performance. The constant calculation: How am I coming across? Did that answer land? Are they bored? Should I wrap up?
Human interviewers—even good ones—create social pressure just by being there. Candidates mirror energy, adjust tone, fill silences nervously. They optimize for rapport, sometimes at the expense of substance.
Remove the human from the screening stage, and some of that pressure evaporates. Candidates still want to do well, but they’re not managing a relationship at the same time. They’re just responding to questions.
The result, paradoxically, is often more honesty. More complete answers. More of the person, less of the performance.
What does change
I’m not going to pretend nothing shifts when candidates know it’s AI. Some things do.
A small percentage of candidates disengage. They give shorter answers, treat it like a form to get through. This is real, and it’s a signal in itself—though you have to be careful not to over-index on it. Some people just prefer human interaction. That’s legitimate.
Others try to optimize. They Google “how to beat AI interviews” and find advice about keywords, pacing, hitting certain phrases. Most of this advice is either wrong or based on outdated systems. Modern voice AI isn’t scanning for keyword density—it’s evaluating how someone structures a response, explains their thinking, handles a scenario.
But yes, some candidates approach it more strategically than they would a human conversation. Whether that’s a problem depends on your perspective. Strategic communication is a skill too.
The candidates who opt out
Let’s talk about the ones who decline entirely.
Some candidates, when told the first interview is AI-conducted, withdraw from the process. This happens. It’s a small percentage in our experience, but it’s not zero.
What do you lose? Potentially some strong candidates who have a genuine preference for human interaction. People who find the idea of talking to AI uncomfortable or who’ve had bad experiences with clunky automated systems in the past.
What do you filter out? Also some candidates who weren’t that interested to begin with and used the AI disclosure as a convenient exit ramp.
The honest answer is that you can’t always tell which is which. If you’re hiring for a role where human connection is paramount—say, executive recruiting or high-touch client service—you might weigh this trade-off differently than if you’re screening hundreds of applicants for an SDR class.
Making disclosure work
How you disclose matters almost as much as whether you disclose.
A cold, legalistic notice buried in an email footer creates a different impression than a clear, friendly heads-up that explains the why. Candidates respond better when they understand the purpose: We use AI for initial screens so every applicant gets a consistent, fair interview—and so our recruiters can spend more time with candidates who move forward.
Frame it as a feature, not a footnote.
Also: give candidates a chance to ask questions or flag concerns before they begin. Some will have accessibility needs. Some will have legitimate technical constraints. A human fallback option—even if rarely used—signals that you haven’t fully automated your way out of caring.
What we’ve learned at RoundOne
We’re transparent about what we are. Every candidate who interacts with RoundOne knows they’re talking to AI. It’s in the invite, it’s in the interface, it’s in the conversation itself.
What we’ve found is that most candidates adapt quickly. After the first question or two, they stop thinking about the medium and start thinking about their answers. The format fades into the background.
The feedback we hear most often isn’t “that was weird” or “I hated talking to a bot.” It’s “that was actually less stressful than I expected” and “I felt like I could take my time.”
Not everyone loves it. But more people are comfortable with it than the discourse would suggest.
The takeaway
Disclosure isn’t a liability. Done well, it can actually improve the candidate experience—less pressure, more consistency, a clearer sense of what to expect.
The candidates who thrive in AI screens aren’t gaming the system. They’re the ones who were going to give thoughtful, structured answers anyway. The format just gets out of their way.
And the ones who struggle? Often, they’d have struggled in a human screen too. You’re just seeing it earlier.
Want to see how candidates actually respond? Try RoundOne free for 14 days—no credit card, no pressure.