Almost everyone job hunting is using AI now. SHRM’s research puts it at about 74% of job seekers reaching for it somewhere in the process, which means the interesting question stopped being whether you should and became where it actually helps versus where it quietly works against you.
The part that’s just smart
Most of what people do with AI in a job search is fine, and honestly, useful. Researching a company before you talk to them, getting a read on the team and the product, drafting a first pass at a cover letter you’ll then rewrite in your own voice: none of that is a gray area. It’s the modern version of doing your homework, and skipping it just means showing up less prepared than the person after you.
The same goes for your resume. Pasting in a job description and asking where your experience lines up, or where it obviously doesn’t, is a good use of ten minutes. It catches the mismatch between what you did and how you described it, which is a common reason solid candidates get filtered out. Used this way, AI is a research assistant and an editor, and there’s nothing to feel weird about.
Practice is where it earns its place. You can run through likely questions, get a sense of which of your answers ramble, and tighten them before they matter. Practicing a full conversation before the real one, against something that actually asks a follow-up instead of accepting your first pass, is how the good examples start coming out cleanly under pressure. That’s the openskill pitch in one line, and I’ll leave it there, because the point stands whether or not you use us for it.
The part that backfires
Then there’s the other 22%. SHRM found that roughly that many job seekers admit to using AI live, in real time, during actual interviews, reading generated answers off a second screen while the interviewer talks. I understand the temptation, especially when the market feels this tight. But it tends to fail, and it fails in ways that are hard to recover from.
The tells are consistent. Answers fed in real time come out oddly paced, a beat behind the question, and they’re generic in a way that stands out precisely because everyone else is also using the same tools. Worse, they collapse on contact with a follow-up. An interviewer asks “what would you have done differently?” or “walk me through the part that broke,” and there’s nothing underneath the polished first answer to draw on, because you didn’t actually live it. Interviewers have gotten good at this. A polished answer that can’t survive one specific question is a bright red flag now.
The part that breaks trust
Further out on the same spectrum is the outright fakery, and this one’s escalated fast. By the middle of 2025, about 17% of HR professionals reported encountering a deepfake in a video interview, someone using synthetic video or a stand-in to misrepresent who was actually applying. That’s a small slice, but it’s large enough that hiring teams now build detection into their process, which makes life harder for every honest candidate on a video call.
What gets lost in the arms race, though, is that even when the fakery works and you get the offer, you’ve won a job you can’t do, or can’t do the way you claimed. The gap doesn’t close when you sign. It shows up in the first week, in front of the people who now have every reason to distrust you. There’s no version of that trade that comes out ahead.
Building the muscle instead of the crutch
The pattern across all of this is pretty clear. AI used before the interview, to prepare, tends to make you better. AI used during the interview, to perform, tends to expose you. And AI used to impersonate ability you don’t have just moves the reckoning to a worse moment.
The durable move is the least clever one, which is to actually get good at the thing. Do the research. Pressure-test your resume. Practice a real conversation enough times that your best examples come out without a script, so that when an interviewer pushes on the hard part, you have somewhere to go. Do that and you never need the second screen, because the answers are already yours.
The honest read
The line comes down to whether you’re using a tool to prepare or to pretend, which has less to do with AI than it sounds. Preparation compounds, it makes the next interview easier and the offer more likely to fit. Pretending borrows against a future you’ll have to face down eventually, usually within days of starting. Use the tools to build the muscle, and skip the ones that only paper over the fact that you didn’t.