"AI is rejecting 90% of resumes" is the kind of stat that spreads because it's terrifying, not because it's precise. Here's what's actually happening in most of those cases: the resume never got read at all. Not because an algorithm judged it and said no, but because the parsing software behind the ATS couldn't turn the file into structured data in the first place. A resume that fails to parse cleanly doesn't get a low score. It gets scrambled, or partially dropped, before anyone or anything ever evaluates it.
This isn't a "how ATS works" explainer. If you want that, we've covered it here: how applicant tracking systems read your resume. This is a line-item breakdown of what actually breaks during parsing, based on the specific formatting choices that consistently cause problems, so you can check your own resume against each one.
The "90% Rejection" Myth vs. What's Really Happening
Most of what gets called an "AI rejection" is really a parsing failure. The system didn't weigh your experience and decide you weren't qualified. It tried to extract your job titles, dates, and skills into a structured record, hit a layout it couldn't interpret, and produced a broken or incomplete version of your resume for a recruiter or hiring manager to see later, if it even got that far.
That distinction matters because the fix is completely different. A judgment problem means rewriting your experience. A parsing problem means changing your formatting. Most of what follows is the second kind, and it's fixable in an afternoon.
The Parsing Failures That Actually Break Resumes
These are the formatting choices that most consistently cause ATS software to misread, scramble, or drop parts of a resume.
Multi-column layouts and text boxes
Most parsers read a page left to right, top to bottom, the way a scanner reads plain text. A two-column layout with your skills in a sidebar and your experience in the main column doesn't read as two separate sections to a parser. It often reads as one blended, interleaved mess, with a line from your skills list landing in the middle of a job description.
For example, a resume with "Python, SQL, Excel" in a left sidebar next to "Managed a team of 6 analysts" in the main column can get extracted as a single garbled line: "Python, SQL, Managed Excel a team of 6 analysts." Nothing about the content changed, but the order the parser reads it in did, and that's often enough to make a strong bullet point unreadable.
Text boxes cause a related problem: some parsers skip their contents entirely, treating them as a separate object rather than part of the page's main text flow. A summary or contact block placed inside a text box for visual polish may never be extracted at all, even though it looks perfectly normal on screen.
Tables used for layout
Tables built for visual formatting, rather than for genuinely tabular data like a list of certifications with dates, tend to confuse parsers about what belongs together. A job title in one cell and its dates in an adjacent cell can get separated during extraction, so the parser ends up with a title and no date, or a date with no job attached to it. Once that link breaks, the parser has no reliable way to calculate how long you held that role, which matters for any system checking years of experience against a job requirement.
Headers and footers
Contact information placed in a page header or footer is one of the most common silent failures, because it's invisible in the way that matters most: the candidate never notices it's a problem. Many parsers don't scan headers and footers at all, since that space is built for things like page numbers or document titles, not core content. If your name, phone number, or email lives there, it may never make it into the parsed record, even though it's clearly visible to a human reading the PDF and looks perfectly professional.
This is one of the more common places resumes fail without the applicant ever knowing why, since everything looks fine on the page.
Non-standard fonts and embedded graphics or icons
Decorative fonts, skill-level graphics (dots, bars, or icons meant to show proficiency), and embedded images can all cause extraction problems. Some parsers rely on standard font encoding to read text correctly, and unusual or heavily stylized fonts can produce garbled or missing characters during extraction. Icons and graphics generally don't get read as content at all, since a parser is looking for text, not images.
If a skill only exists as a filled-in bar or a small icon next to a label, with no accompanying text like "Advanced" or "3 years," that information can disappear entirely from the parsed version of your resume.
Section headers that aren't recognized
Parsers are typically trained to recognize a defined set of standard section labels: "Experience," "Education," "Skills," and their close variants. A creative header like "Where I've Been" instead of "Experience," or "What I Bring" instead of "Skills," can mean the parser fails to identify that block as work history or skills at all. Depending on the system, that content can end up dropped into an unclassified section that never gets weighted the way a properly labeled section would.
Creative section headers might stand out to a human reader skimming quickly, but that same creativity is exactly what confuses a system trained on standard terminology.
Date formatting inconsistencies
Parsers try to build a timeline from your work history, both to check for continuous experience and to calculate tenure in each role. Inconsistent date formats across entries, mixing "Jan 2022," "01/2022," and "2022" on the same resume, can prevent that timeline from assembling correctly. In some cases this can make real, continuous experience look like it has a gap, simply because the parser couldn't confidently match a start date to an end date.
File format issues
A resume saved as an image-based PDF (essentially a picture of your resume rather than a text-based document with selectable text) can't be parsed at all in some systems, since there's no underlying text layer to extract. This tends to happen with resumes exported from certain design tools built for visual layouts rather than documents, or with scanned paper resumes. If you can't highlight and copy text directly from your own PDF, there's a real chance an ATS can't read it either.
Quick Reference: What to Avoid
- Multi-column layouts or text boxes for your summary, skills, or contact information
- Tables used to lay out job titles, dates, or companies
- Contact information placed in a header or footer instead of the main body
- Decorative fonts, skill-rating graphics, or icons used in place of written text
- Non-standard section headers instead of "Experience," "Education," and "Skills"
- Inconsistent date formats across different entries
- Image-based PDF exports where the text can't be selected or copied
A Quick Example: Same Content, Two Different Results
Picture a resume with a two-column layout: a narrow left sidebar listing "Skills: Excel, SQL, Salesforce" and "Certifications: PMP, Six Sigma," next to a wider right column with the actual job history, starting with "Marketing Manager, Acme Co., 2021-Present: Led a team of 5, grew email revenue 30%."
To a person looking at the page, that's a clean, professional layout. To a parser reading left to right, top to bottom, ignoring the visual columns, it can come out as something closer to: "Skills: Excel, SQL, Marketing Manager, Salesforce, Acme Co., 2021-Present: Certifications: PMP, Led a team of 5, Six Sigma, grew email revenue 30%."
The job title, company, and accomplishment are all still technically present, but they're interleaved with unrelated skills and certifications in a way that makes them difficult or impossible to reliably extract as a single coherent job entry. Rebuild the same resume as a single column, with skills and certifications in their own section below the work history instead of beside it, and the exact same content extracts cleanly, in order, with nothing lost.
That's the core of what this entire article is about: the words on the page didn't change. Only the layout did, and that alone was the difference between a resume that reads correctly and one that doesn't.
How to Tell If Your Resume Has One of These Problems
A quick, low-effort check: copy the text out of your resume file and paste it into a plain text editor, like Notepad or TextEdit, with formatting stripped out. This won't perfectly replicate any specific ATS, but it's a close approximation of what a parser is working with once it strips away your visual design.
Look for a few specific things once you paste it in:
- Is your contact information present at all, or did it disappear because it was in a header or footer?
- Are your skills and job history interleaved or scrambled together, rather than appearing as separate, readable blocks?
- Does your work history appear in a logical order, with each job's title, company, and dates staying together?
- Are there strange character strings or missing words where an icon, graphic, or unusual font symbol used to be?
- Do your section headers still read as "Experience," "Education," and "Skills," or did creative labels get lost in the shuffle?
If you see any of these, it's a strong signal that the same problem is likely happening inside an actual ATS, just without the visual cue of a garbled text file to warn you.
If you'd rather not do this manually, or want a second confirmation, check your resume for free with our ATS score checker, which flags formatting issues like these alongside your overall score.
What Actually Gets You Past the Parser
The fixes mirror the problems: a single-column layout, standard section headers, contact information in the body of the document rather than a header or footer, a standard font, and a text-based (not image-based) PDF export. None of this requires sacrificing a resume that also looks good to a human reader. Clean, simple formatting reads well for both a parser and a person, which is part of why it's the safer default rather than a compromise.
It's worth separating this from keyword matching, which is a different problem entirely. Fixing your formatting means the parser can correctly extract what's already on your resume. It doesn't address whether the content itself uses the language a specific job posting is looking for. If you've confirmed your formatting is clean and you're still not hearing back, the next thing worth checking is whether your resume's actual wording matches the posting's, which is a separate step covered here: how to find keywords in a job description.
If you want to see how different tools score the same resume, including which ones catch formatting issues like these and which don't, we put seven of them through an identical test here: we tested 7 ATS checkers on the same resume.
Why So Many Resumes Use Risky Formatting in the First Place
Most of the formatting choices covered here aren't mistakes in the sense of carelessness. They're usually the result of using a template built to look impressive to a human reader first, with no consideration for how it gets processed by software. Design-forward resume builders and templates often use columns, graphics, and creative section headers specifically because they stand out visually in a stack of PDFs. That's a reasonable goal, and those templates aren't wrong to exist. The problem is that visual polish and machine readability are two different design goals, and a template optimized for one isn't automatically safe for the other.
This is also why the same resume can perform inconsistently across different applications. A simpler, older-fashioned template might get read correctly everywhere, while a modern, visually striking one might work fine at one company and silently fail at another, depending on which ATS that employer happens to use. The inconsistency itself is often the biggest clue that a parsing issue, not a qualifications issue, is at play.
Check Your Resume Before You Apply Again
Rather than guessing which of these issues might apply to your resume, or manually pasting it into a text editor, you can see exactly how it scores and what it flags in under a minute.
Related reading
If your formatting checks out and you're still not hearing back, the next place to look is whether your resume's actual wording matches what a specific job posting is scanning for: how to find keywords in a job description. And if you want to see how far scores can vary tool to tool on the exact same resume, we tested that directly: we tested 7 ATS checkers on the same resume.
Frequently Asked Questions
Is it true that AI is rejecting 90% of resumes? That figure circulates widely, but the more accurate description is that a large share of resumes never parse correctly into structured data in the first place. That's a formatting failure, not a qualification judgment, and it's usually fixable.
Can a well-written resume still fail an ATS? Yes. Parsing happens before content is evaluated. A resume with strong, relevant experience can still be misread or partially dropped if it's built with a layout the parser can't interpret correctly.
Do all companies use the same ATS? No. Different platforms parse documents differently, and some handle complex formatting better than others. Sticking to simple, standard formatting is the safest approach because it holds up across the widest range of systems.
Will fixing formatting guarantee an interview? No. Fixing parsing issues gives your actual experience a fair chance to be read and evaluated. It doesn't change how qualified you are for a role, only whether your real qualifications make it into the system correctly.
Related reading: Once your resume passes ATS parsing, the next step is tailoring it to the specific job description, see How to Tailor Your Resume for Every Job in Under 10 Minutes.