Why your résumé gets rejected
Almost everything written about this is built on one number that turns out not to have a study behind it. Here is what actually happens to an application, in the order it happens, and which parts a résumé can do anything about.
First: the “75% are rejected by a robot” number
It traces to a résumé-optimisation vendor’s sales material from 2012. The vendor was gone by 2013. There is no study. It has been repeated for over a decade by companies whose product only makes sense if you believe it, and it is repeated most often by the ones selling you a score to raise.
What recruiter surveys actually show is content-based automatic rejection well under 10% of configurations, and the major hiring platforms do not reject on résumé content by default. The filters are real. They are just not the one you were sold.
What actually happens, in order
- 1
The knockout questions on the form
Before anything reads your résumé, the application form asks a handful of questions the employer has configured to disqualify outright: are you authorised to work here, will you need sponsorship, can you work from this location, do you hold this licence. Answer one of them the wrong way and the application is closed by a rule, immediately, whatever the résumé says.
This is the only genuine automatic rejection in most hiring systems, and it is the one nobody optimises for, because every product in this category is selling résumé advice.
Read the questions before you write anything. If a posting requires something you do not have, the honest move is to skip it and spend the hour on one you do — not to answer around it, which is a false statement on a form with your name on it.
The scanner cannot measure this — it is the application form, not the résumé.
- 2
A file the software cannot read
What gets stored is not your PDF. It is whatever a parser recovered from the text layer: a name, an email, a list of roles with dates. A résumé exported as an image has no text layer at all, so it recovers nothing. A two-column layout is read as one top-to-bottom stream, so the columns interleave. Content in a header or footer is frequently not read at all.
You do not get rejected for this. You become invisible to it — recruiters search the stored record, and a record with no skills in it is not in the results. Measured parse rates run around 93% for single-column layouts and roughly 86% for multi-column ones.
One column. Standard headings. Contact details in the body of the page, not the header. Export to PDF rather than printing to an image, and check that you can select the text.
The scanner measures this one.
- 3
A page that claims more than it shows
Once a person is reading, the question stops being whether the words are present and becomes whether anything is behind them. A skill sitting alone in a comma-separated list is an assertion. The same skill inside a bullet that describes what was built and what changed is evidence.
This is where a long skills line stops helping and starts hurting. A reader who notices that twenty of your thirty listed technologies appear nowhere in your work does not conclude you are versatile.
For each skill on the list, find the bullet that shows it. If there is not one, either write the bullet or cut the skill.
The scanner measures this one.
- 4
Seven seconds of triage
The first human pass is not reading. It is scanning verbs, numbers, and the first two bullets of the newest role, deciding whether to read properly.
A résumé that survives the machine and then opens with "Responsible for managing the team" has spent its seven seconds saying what the job title already said.
Open each bullet with what you did and close it with what changed, in a number. Put the strongest one first in the newest role.
The scanner measures this one.
Why we do not give you one number
In August 2026 we ran two résumé products’ real output through a competitor’s own ATS score. A heavily fabricated résumé came back at 89 / 100, “Excellent”. The same product scored 35 against a different job. We followed those fabricated résumés into 21 real applications and they produced zero interviews.
A score that rises when you add keywords cannot tell a real accomplishment from an invented one, so it rewards exactly the thing that does not work. We report two numbers instead — what a job asks for that you have, and how much of that a reader can actually find you doing — and the distance between them.
See it on your own résumé
Stages two, three and four are all measurable from the file itself. Upload it and the scanner names the specific lines rather than giving you a grade.
Scan my résumé