Last year a hiring manager at a mid-size fintech told me she’d posted a role, set up automated screening, and then didn’t look at the applicant pool for nine days. By the time she logged back in, 340 applications had been filtered down to 23. She offered the role to someone from that pool of 23. All 317 others got a rejection email from software.
That’s roughly 93% of applicants filtered before a person read a single word. The number doesn’t shock me anymore, but it should.
How ATS filtering actually works (it’s less smart than you think)
Most applicant tracking systems aren’t doing semantic analysis of your experience. They’re doing keyword matching, title normalization, and sometimes years-of-experience parsing. The more sophisticated systems from Workday or Greenhouse have gotten better at contextual reading, but “better” is relative. They’re still fundamentally matching text against a rubric the recruiter set up, often quickly, often imperfectly.
The practical implication: if your resume says “overseeing revenue operations” and the job posting says “managing revenue operations,” some systems will count those as different terms. I’m not making that up. Using the exact phrases from the job posting matters more than you’d expect from a system that supposedly understands language.
According to Jobscan’s analysis of major ATS platforms, resumes that include the exact job title from the posting are around 10 times more likely to surface in recruiter searches. That’s not a guarantee of an interview, but it’s a significant filter to clear.
Formatting rules that are actually worth following
Some ATS formatting advice you’ll find online is outdated or overly cautious. Here’s what I think still applies in 2026.
- Use standard section headers. “Work Experience” parses better than “Where I’ve Been.” No joke, I’ve seen candidates do this.
- Don’t put anything critical in a header or footer in Word/PDF. Some parsers skip those zones entirely.
- Tables and columns in Word format can scramble the order of content when parsed. If your resume has a two-column layout, test it by copying the text out and seeing if it reads coherently in plain order.
- PDFs are usually fine now. The “never submit PDFs” advice is at least four years stale for most major platforms. But if the application portal asks specifically for Word, give them Word.
Avoid custom icons, graphics, and sidebars with color blocks. Not because they look unprofessional to a human, but because a parser that hits a graphic element mid-scan sometimes drops everything that follows it. That’s a real, documented failure mode.
Keywords: the part most people underdo
The actual work is matching your resume’s language to the specific job posting you’re applying to. Not once with a template and then mass-applying. Per application, or at minimum per role type.
Take the job description, paste it into a text file, and find the six to nine phrases that appear most frequently or seem most central to the role. Then look at your resume and ask: are those phrases there? In the work experience section, not just a skills list at the bottom? If not, find places where you can accurately use them.
“Accurately” matters here. Don’t stuff keywords for phrases that don’t genuinely describe your work. ATS gets you past the filter; the hiring manager reading your resume on the other side will catch anything that doesn’t hold up in an interview.
Include both acronyms and full versions where relevant. Write “AWS (Amazon Web Services)” at least once. Some systems normalize these; some don’t. Takes you thirty seconds and removes the uncertainty.
The skills section question
Most ATS guidance tells you to put a skills section at the top of your resume. I think this is slightly wrong for experienced candidates. A skills list divorced from context (“Python, SQL, Project Management, Stakeholder Communication”) tells the system a lot and tells a human almost nothing.
Better approach: put the most important keywords in your actual experience descriptions. “Led a team of seven engineers to build a Python-based data pipeline processing 4M daily records” beats “Python” in a skills list in almost every way, for both parsers that understand context and humans who will read it after.
Keep a condensed skills section for quick scanning, but don’t treat it as the primary keyword vehicle.
What “ATS optimization” doesn’t fix
A well-formatted, keyword-matched resume gets you into the pool of candidates a human reviews. It doesn’t get you an offer. It doesn’t compensate for a resume that shows thin experience for a senior role, unexplained gaps that raise questions, or a career narrative that doesn’t obviously connect to the job.
The Bureau of Labor Statistics Occupational Outlook Handbook shows that hiring in most professional and technical fields is still net positive through 2026, which means there are real roles being filled. The competition is real too. Clearing the ATS filter matters, but getting past the human screen after that is a separate problem with different solutions.
Where Craqly can help is on that second part: practicing your story for phone screens and interviews so you can actually explain what’s on the resume when someone asks. Getting through the filter and then fumbling the recruiter call is more common than people admit.
One thing I genuinely don’t know
I can’t tell you with confidence how much AI-assisted resume scoring (as opposed to basic keyword matching) has changed actual outcomes in 2026. The vendors all claim their AI screening is more accurate and fair. Third-party audits of those claims are sparse. I’d proceed as if keyword matching still dominates, because even where AI scoring is in use, the training data for those models came from keyword-filtered historical hiring. The biases stack.
Clear the filter. Then make your case to a person. Those are two different jobs.