A friend of mine applied to eleven jobs last spring and got exactly one human response. Not one rejection, one response, as in a real person acknowledging she existed somewhere in the process. The other ten either vanished into silence or got filtered out by something before a human ever opened her resume. She’s a strong candidate too, the kind of person who used to get interviews without much trouble. Something’s clearly different now.
Resume Screening Isn’t Really Done by Humans Anymore
Most mid-size and large companies run resumes through some automated screening before a person ever looks at them. That’s not new exactly; applicant tracking systems have existed for years, but the AI layered on top of them now does a lot more than keyword matching.
What Changed From the Old Keyword Filters
The old systems were dumb in a predictable way. Miss the exact phrase “project management” on a resume and get filtered out, even if every bullet point described project management work using different words. Annoying, but at least somewhat gameable once people figured out the trick.
The newer AI-driven versions are harder to reverse-engineer. They’re trained to recognize skills described in different language, which sounds like an improvement, and often is, but it also means the screening is less transparent. My friend’s resume got rewritten three separate times trying to figure out what was tripping the filter, and she still isn’t totally sure which version actually worked, because the one job that did respond was one she applied to without any changes at all.
Where This Actually Helps Candidates
It’s not all bad news, to be fair. A recruiter I talked to who works in-house at a mid-size tech company said the AI screening genuinely catches qualified candidates that old keyword systems used to miss entirely, people who described their experience in unconventional language or came from a non-traditional background. She said her team’s candidate pool got noticeably more diverse once they switched systems, which wasn’t really the pitch when they bought the software, more of a side effect nobody specifically asked for but was happy to see.
Interviews Are Getting the AI Treatment Too
Screening resumes was just the start. A lot of companies now use AI somewhere in the actual interview process itself, and this is the part that makes people genuinely uneasy.
One-Way Video Interviews
The format that comes up most often is the one-way video interview, where a candidate records themselves answering a set of questions with no human on the other end, at least not in real time. An AI system reviews the recording afterward, sometimes analyzing word choice and speech patterns, occasionally facial expressions too, depending on the vendor.
My friend did one of these and described it as talking into a void. No reactions, no follow-up questions, no sense of whether an answer landed or missed completely. She said it felt less like an interview and more like filming an audition tape for a part she didn’t fully understand.
The Accuracy Question Nobody Has Fully Answered
Whether these systems actually predict job performance well is genuinely unresolved. Some vendors claim strong validation data. Independent researchers have raised real concerns about bias creeping into these systems, particularly around accent, speech patterns, and anything that deviates from whatever dataset the tool was originally trained on. A system trained mostly on one demographic’s interview patterns doesn’t magically generalize fairly to everyone else, and there have already been a handful of public cases where companies quietly pulled back on AI interview tools after discovering exactly that kind of skew.
Candidates Are Using AI Right Back
Here’s the part that complicates the whole picture even further. While companies lean on AI to screen candidates, candidates have started leaning on AI to get past the screening, and it’s turned into a quiet arms race nobody quite admits to openly.
Resume Writing Has Changed
Tailoring a resume to match a specific job posting used to take real effort. Now it’s a five-minute task for anyone willing to run their resume and the job description through an AI tool and ask for suggested edits. My friend eventually did exactly this, somewhat reluctantly, after her manually tweaked versions kept failing to get responses.
Coaching Through the Interview Itself
Some candidates have gone further, using AI tools in real time during interviews, reading suggested answers off a second screen during a video call. Recruiters have started noticing the telltale signs, slightly delayed responses, eyes flicking somewhere off-camera, answers that sound oddly polished and strangely similar across different candidates. One hiring manager told me she’s started asking more unpredictable follow-up questions specifically to catch this, since a scripted or AI-generated answer tends to fall apart fast once a conversation veers off the expected path.
What This Means If You’re Actually Job Hunting Right Now
For anyone currently in the middle of this, the practical takeaway is a little uncomfortable but worth saying plainly. Resumes need to describe actual skills in plain, varied language rather than chasing exact keyword matches, since the systems reading them are looking for meaning now, not just string matches. Leaning on AI tools for first drafts is fine and increasingly normal, but genuinely knowing the material well enough to answer unscripted follow-up questions matters more than ever, precisely because recruiters are actively watching for the gap between a polished prepared answer and real, in-the-moment understanding.
The Human Layer Still Decides Most of the Time
Despite all of this, nearly everyone I talked to while looking into this, recruiters and candidates both, agreed on one thing: a human still makes the final call in the overwhelming majority of hiring decisions. AI narrows the pool and flags people for a closer look. It’s rarely the thing actually deciding who gets the offer. That’s at least somewhat reassuring, even if the path to get in front of that human has gotten noticeably more complicated than it used to be.
Small Companies Are Catching Up Fast
It’s worth mentioning that this isn’t just a big-company phenomenon anymore either. A smaller business I know of, maybe thirty employees total, started using an AI screening tool last year mostly because the owner got tired of manually reading two hundred resumes for a single opening. She told me she felt a little guilty about it at first, like she was outsourcing something that should’ve stayed personal. That feeling faded fast once she saw how much time it actually saved, though she still insists on personally reading every resume that makes it past the first filter, which feels like a reasonable middle ground between the old way and the fully automated one.
Where This Leaves Things
My friend eventually got a job, for what it’s worth, through a referral rather than any of the systems described here, which is its own kind of answer about how much of this is really working the way it’s supposed to. The tools aren’t going anywhere. They’re just one more layer candidates now have to understand and work around, on top of everything hiring already asked of people before AI showed up and made the whole process a little stranger.