Are You Interviewing the Candidate—or Their AI?

For years, the job interview has been one of the most important parts of the hiring process. A resume tells you where someone has worked and what they claim to have accomplished, but the interview is where you actually meet the person behind it. You hear how they communicate, ask about their experience, see how they respond when you go off-script, and get a sense of whether they’re really the right fit.

But with the rapid advancement of artificial intelligence, I think we need to start asking an uncomfortable question: can we still assume the person answering our interview questions is actually demonstrating their own knowledge and abilities?

Recent developments suggest the answer is more complicated than we’d like.

A Candidate Who Didn’t Exist

A recent HRD Canada article about an experiment involving a completely fictional job candidate caught my attention. Cybersecurity advisor Jake Moore built an entire professional identity around a fictional candidate named Jackie Morris—resume, LinkedIn profile, Instagram account, passport, AI-generated appearance, altered voice. He then applied for a real marketing position.

Out of 262 applicants, four were invited to interview. Jackie Morris was one of them. She made it through the process and was offered the job.  The problem? Jackie Morris didn’t exist.

What makes this even more concerning is that the deception didn’t stop at a convincing resume and online profile. During the live interview, Moore ran a program in the background that listened in real time and generated suggested answers whenever he wasn’t sure how to respond. The interviewer came away impressed by how knowledgeable the candidate seemed.

A fictional candidate, complete with a fabricated professional identity, made it through multiple stages of a hiring process and convinced an employer she was a qualified person. Think about that for a moment.

When AI Starts Answering the Interview Questions

We’re also seeing something that may be even more relevant to everyday hiring: software built to listen to an interviewer’s questions and feed candidates suggested answers in real time, designed to run discreetly during virtual interviews.

Most people don’t have a problem with candidates using AI to prepare. If someone uses it to research a company, practice interview questions or sharpen how they talk about their experience, that’s not dishonest. People have always prepared, whether with friends, career coaches or their own notes.

The problem starts when AI moves from helping someone prepare to supplying the answers during the assessment itself.  If I ask a candidate how they handled a difficult customer, I want to hear about their experience—what happened, what they did, why, and what they learned. If an AI tool is generating that answer for them in real time, I’m evaluating their ability to use AI, not their ability to do the job.

A Polished Interview Doesn’t Always Mean a Qualified Candidate

Anyone who’s spent years interviewing candidates knows a polished interview doesn’t guarantee success in the role. There’s nothing new here.  The strongest candidates usually aren’t the ones with the most rehearsed answers—they’re the ones who can have a genuine conversation with you. They tell you what actually happened, why they made a particular call, what happened next. When you ask for further details, they don’t need to search for another perfect answer, because they experienced it themselves.

I’ve always believed in going beyond standard interview questions—using a candidate’s real experiences and authentic stories rather than relying on rehearsed responses. You get much more out of the person when asking strategic questions.

Ask them, “Tell me about a time you dealt with a difficult customer,” and dig deeper. What was the customer upset about? What did you say? How did they respond? What did you do next? Who else was involved? How did it end? Looking back, would you handle it differently today?

The goal here isn’t to “catch” someone using AI—it’s to have a genuine conversation that lets you understand whether they actually lived the experience they’re describing. The more conversational the interview becomes, the more you’re assessing the person instead of a perfectly prepared answer.

The Problem Goes Beyond AI-Generated Answers

AI can now touch almost every stage of how a candidate presents themselves—the resume, the LinkedIn profile, prepared interview answers, real-time interview assistance, and in some cases, a manipulated voice or appearance.

I flagged the deepfake side of this over a year ago, when live interviews were already starting to show fabricated personas and employers needed to rethink how they verify candidates. The technology has evolved since. The Jackie Morris experiment shows just how far it’s come: a convincing professional identity built with relatively inexpensive, commercially available tools, combined with real-time AI assistance during the interview itself. Other recent reports point to AI-generated voices playing a role in candidates actually getting hired.

Things are moving quickly enough that employers can’t build one checklist of “AI red flags” and assume it will protect them for long.

I don’t believe we need to stop conducting virtual interviews, and I don’t think we should assume every candidate with a polished answer is cheating. I think we need to get better at interviewing and at telling the fake answers from the real ones.

So what should employers do?

Ask the same questions both virtually and in person:  That means more follow-up questions. We’ve done this ourselves for years, even before AI was a thing. We ask the same questions over the phone, in person, and sometimes again during a follow-up call. It’s helped us catch dishonest candidates and confirm consistency in the answers we’re getting. More than once, we’ve watched a candidate struggle to connect the dots and keep their story straight.

Go deeper with candidates:  Asking candidates to explain how they reached a decision, and posing the same questions more than once, virtually and in person. We need to pay attention to the whole conversation. Does their experience add up? Is it consistent with what they told you earlier in the process? Can they explain the details of work they say they performed? Can they handle an unexpected follow-up? Do they actually understand the terminology they’re using?

None of this is complicated. It’s just good recruitment practice—and in my experience, it tells you a lot.

This matters even more when you’re hiring bilingual talent.

At BlueSky Personnel Solutions, we’ve specialized in English/French recruitment for more than 25 years, and one thing we know well is the gap between saying you’re bilingual and actually being able to work professionally in both languages.

AI makes that gap easier to hide. Someone can ask AI to draft a polished French answer, translate a response, and dress up the vocabulary until it sounds thoroughly professional. But can they actually hold a conversation in French? Can you switch from English to French mid-interview and have them follow along naturally? Can they respond to something unexpected, explain a complicated situation without translating word-for-word, catch nuance, or talk comfortably with a French-speaking customer?

We see this regularly in our own process. A candidate may list themselves as bilingual on paper, but once we start speaking French with them, their actual level can look very different from what the resume suggested. That’s why we don’t take anyone’s word for it—we have a conversation with them right away, and we give them no warning whatsoever. If you say you’re fluent in French, we just naturally switch to French to see if you can answer us on the spot. I think that principle is becoming essential across recruitment generally, not just bilingual hiring.

Employers shouldn’t treat every candidate as a suspect.

Verify important information, and ask the important questions two, three or four times if necessary. Check references and validate credentials where it makes sense. Compare what a candidate tells you at different stages of the process. Ask questions that require them to explain their own experience in their own words. Ask them to dig deeper, especially when they’re in front of you, in person.

I’m not anti-AI. It can genuinely make recruitment better—handling administrative work, scheduling, research and other repetitive tasks, and freeing recruiters and HR professionals to spend more time on the things that actually require human judgment. But recruitment isn’t only about efficiency. At the end of the day, we’re making a decision about a person—whether they have the experience, skills, communication ability, personality and motivation to succeed in a particular organization. That takes judgment. It takes listening. And sometimes it takes asking the question you didn’t plan to ask.

After more than 25 years in recruitment, some of the most valuable things I’ve learned about a candidate came from the conversation that happened after the official interview question was answered. The follow-up. The unexpected scenario. The moment you ask someone to explain something a different way, and how they interact with you in general. Those moments aren’t on a resume—they’re real-life experience and evidence you can’t dismiss.

The traditional hiring process leans on a few familiar signals: the resume, the LinkedIn profile, the interview, the reference check. Those still matter, but they deserve a critical eye. Is the resume accurate? Is the person who submitted it the person we’re speaking with? Are the answers actually coming from the candidate? Can they demonstrate the skills they claim? Can they think when the conversation moves somewhere unexpected? Can they communicate naturally, without AI filling in the gaps?  Ultimately: can they actually do the job?

As AI gets better at creating convincing resumes, convincing answers, and even entire candidates who don’t exist, the human side of recruitment may matter more than ever. When you’re making a hiring decision, you don’t just want the right answer—you want to know whether someone can actually do the job.

And sometimes, the best way to find that out is simply to keep the conversation going.