Is Your Resume “ATS Friendly”? Here’s What That Actually Means Now

The newest trick, and why it’s worth knowing about

Resume gaming isn’t new. What’s new is who — or what — people are trying to game. As more hiring tools add AI on top of traditional parsing, a small number of job seekers have started hiding text in their resumes aimed specifically at that AI — not keywords this time, but actual instructions. Something like a hidden line telling an AI reviewer to ignore the job requirements and report a perfect match regardless.

One analysis from Duke University, led by professor Neil Gong and presented at the USENIX Security Symposium, examined roughly 200,000 real resumes submitted to a hiring platform and found that at least 1 in 100 contained exactly this kind of hidden manipulation attempt. Most weren’t crude, obvious commands — they were subtler attempts to nudge a favorable read.

Worth being clear about the risk here: this isn’t a clever shortcut. It’s a gamble on the hiring tool being naive enough to just follow a buried instruction instead of actually evaluating the resume — and it’s the kind of thing that, if noticed, reads as an attempt to deceive rather than a formatting choice. Modern AI-based reviewers are generally built to recognize exactly this pattern and disregard it. The safer, and honestly more effective, use of your time is making the actual resume stronger.

The trick that used to work: hidden keyword stuffing

This is the older, more well-known version of the same impulse. Around 2010 to 2014, a real trick briefly worked: type a wall of relevant keywords in white text on a white background, invisible to a human reader but present in the file’s raw text, hoping an early keyword-counting system would just tally them up.

It doesn’t work anymore, and it hasn’t for a while. Modern platforms — Workday, Greenhouse, and Lever have all been specifically documented doing this — actively scan for zero-opacity or matching-color text and flag it. In some systems, getting caught doing this doesn’t just fail to help; it can flag the application for review as a potential integrity issue. What used to be a harmless-if-pointless trick is now a genuine reason to get a second, more skeptical look.

The myth that won’t die: “75% of resumes get auto-rejected”

This number gets repeated constantly, almost always with no source attached. It doesn’t hold up. A 2025 Enhancv study interviewing 25 U.S. recruiters across more than ten major ATS platforms found that 92% said their systems don’t automatically reject resumes based on content, formatting, or keyword matches at all. Only 8% used any form of broad, content-based auto-rejection, and even then, typically only for narrow, specific requirements.

That doesn’t mean ATS software doesn’t matter. It clearly does — sorting and ranking still determines who a human sees first, and in what order. But the popular image of a robot instantly trashing three-quarters of all resumes on sight is mostly fiction. The real failure mode is usually a resume ranking low enough that a busy recruiter never gets to it, not an automatic rejection with no human ever in the loop.

The habit that’s outdated: chasing exact keyword density

For a long time, the safest resume advice was: match the posting’s exact wording, as often as reasonably possible. That’s less true now. A growing share of ATS platforms use semantic matching — meaning “led a team of 5” can register against a posting asking for “team leadership,” even without matching word-for-word.

That’s genuinely useful for job seekers, but it comes with a catch: it rewards resumes that describe real experience clearly, not resumes stuffed with repeated exact phrases. Overloading a bullet point with a keyword five times reads as unnatural to a human reviewer and doesn’t meaningfully outperform saying it clearly once. The better use of that effort is making sure the posting’s important terms appear somewhere in your resume in your own words — not maximizing how many times any single one shows up.

The worry that’s mostly outdated: “ATS can’t read PDFs”

This one used to have some truth to it, back when parsing software was less mature. It doesn’t anymore — modern parsers handle a standard PDF fine, as long as the text is real and selectable, not a scanned image or a flattened design export. A quick way to check: try selecting and copying a line from your own resume PDF. If it copies cleanly, you’re fine. If it doesn’t, that’s the actual thing worth fixing, not the file format.

The part everyone actually wants: a single number

A lot of the searches that lead people to pieces like this one boil down to the same ask, whether it’s phrased as “ATS score checker,” “resume scorer,” or just wanting a number back. That instinct makes sense — a score is easy to compare, easy to act on, easy to feel better or worse about. But it’s worth holding loosely. A single score is only ever a snapshot of how one system, at one moment, reads your resume against one specific posting. Change the posting, and the number changes with it — not because your resume got better or worse, but because the comparison did. Treat any score as a diagnostic pointing at specific gaps, not a verdict on you.

What actually works, still

None of this is really about tricks, old or new. It’s about one plain idea that hasn’t changed: describe your real experience, clearly, in language close to how the posting describes what it wants. That single habit outperforms every shortcut on this list, including the ones that briefly worked and the ones that never did.