Original Data · Updated Live
The ATS Data Report: what actually breaks in real resumes and CVs.
Almost everything written about applicant tracking systems is opinion, or a test run on a handful of sample files. This page is neither. Every figure below is counted from real resumes that real job seekers uploaded to our scanner, and it re-reads itself each time this page loads.
Every failure, ranked
Percentages are of all 2,326 resumes analysed. One resume can carry several issues, so the column does not total 100.
Skills table flattened into glued text
A skills or contact section built as a table collapses into one run of joined words. The words survive, but no individual skill is searchable, so a recruiter filtering for a keyword never matches this resume. How to fix it
Name not detected
The parser could not identify a candidate name at all, usually because it sits in a header, a text box, or a graphic. The application arrives with an empty name field. How to fix it
Em dashes and special characters
Typographic dashes and similar characters survive inconsistently across parsers. When they fail they leave question marks or boxes mid-sentence in the recruiter's view. How to fix it
Icon bullets rendering as junk
Decorative bullets drawn from icon fonts have no plain-text equivalent, so each bullet point can begin with a stray symbol or an empty box. How to fix it
Two-column layout scrambled
The parser reads the file in the order the text was written into it, not the order the eye follows. Sidebar content lands in the middle of the work history, and dates detach from the jobs they belong to. How to fix it
Ambiguous date formats
Mixed or unlabelled date formats leave the parser guessing at employment periods, which is what feeds the years-of-experience filter recruiters sort on. How to fix it
Name merged with other text
The name is found but welded to the job title or phone number beside it, so the candidate record reads as one long unusable string. How to fix it
Smart quotes and curly apostrophes
Word's automatic curly quotes are a different character from the straight ones, and they are a common source of the stray symbols that appear inside otherwise clean text. How to fix it
Missing standard section headings
Creative headings such as 'My Journey' are not recognised as sections, so experience and education may never be mapped to the right fields. How to fix it
Almost no extractable text
The file is effectively an image. A design exported as a picture, or a scanned document, gives the parser nothing at all to read. How to fix it
Contact details stranded in a header or footer
Several parsers ignore headers and footers entirely. When the phone number and email live there, the application arrives with no way to contact the candidate. How to fix it
What the numbers say
The failures at the top of this list are not exotic. They are the default output of ordinary resume templates: a skills table, a name in a header, a decorative bullet, a second column. Nobody who submitted these files did anything careless. They used the tools they were given.
What makes this class of problem worth measuring is that it is invisible on both sides. The applicant sees a clean document and hears nothing back. The recruiter sees a thin or scrambled record and moves on, with no reason to suspect the file rather than the person. There is no bounce message for a resume that parsed badly.
It is also the cheapest category of problem in the whole job search to fix. Rewriting a career story is hard. Moving a phone number out of a header is not.
Method, and what we do not store
- Sample: every resume uploaded to the free scanner since 2 August 2026. It is a self-selected sample of people who suspected a problem, so treat it as the rate among resumes worth checking, not among all resumes on earth. We would rather say that than pretend otherwise.
- Detection: each file is parsed with the text-extraction behaviour documented in our methodology and parsing benchmark, then checked against a fixed list of failure patterns.
- What is stored: one number per failure type, and one number counting scans. Nothing else. No resume text, no file, no name, no email, no country, nothing that could be traced to a person. Files are parsed in memory and deleted within 60 seconds, as our privacy policy says.
- Why it moves: the figures are read live at page load. Refresh after a busy hour and they will have changed. Nothing here is typed in by hand.
- Use it: journalists, researchers and writers are welcome to cite these figures with a link to this page. Always quote the sample size alongside the percentage.
Find out which of these is on your resume.
The scan that produced this data is free, takes seconds, and shows you the exact text a parser pulled out of your file. Your scan becomes one more anonymous row in the numbers above.
Scan my resume, free →Questions about this data
What percentage of resumes fail ATS parsing?
In our live sample of 2,326 real documents, the single most common failure, skills table flattened into glued text, appears in 40.8% of them. Most files carry more than one issue. We deliberately do not publish a single "X% of resumes are rejected" figure, because rejection is a hiring decision made by people and no scanner can measure it. What is measurable is whether the text survives extraction, and that is what every number here reports.
Does this data apply to CVs as well as resumes?
Yes. Every failure listed here is a document-formatting problem, not a regional convention, so a two-column CV breaks in exactly the way a two-column resume does. Our uploads come from both naming conventions: readers in the UK, Ireland, India, Pakistan and the Gulf call the file a CV, readers in the US and Canada call it a resume, and the parser treats them identically.
What is the difference between a CV and a resume for ATS purposes?
For applicant tracking systems, none that matters. The distinction is regional vocabulary plus, in academia, length: an academic CV runs many pages and lists publications, while a resume or a standard UK CV is one to two pages. Parsers do not care what you call the document. They care whether the text can be extracted in a sensible order, which is what this report measures.
What is the most common ATS parsing error?
Skills table flattened into glued text, found in 40.8% of the 2,326 documents scanned so far. A skills or contact section built as a table collapses into one run of joined words. The words survive, but no individual skill is searchable, so a recruiter filtering for a keyword never matches this resume.
Where does this data come from, and how current is it?
From real resumes and CVs that people upload to our free scanner, counted anonymously since 2 August 2026. We store only integer tallies per issue type: no file, no text, no name, no email, nothing that identifies a person or a document. The figures on this page are read fresh on every page load, so they move as scans happen rather than being refreshed monthly.
Can I cite or republish these statistics?
Yes, free of charge, including commercially. The dataset is published under a Creative Commons Attribution licence: cite ATS Verification and link to this page. Journalists and researchers who want a breakdown we have not published can write to hello@atsverification.com.