The ATS resume format, and what happens when a machine reads yours
An ATS friendly resume format is simply a layout a machine can pull clean text out of, in the right order: one column, real text rather than an image, ordinary section headings, and the words the job ad itself uses. That is the whole specification. And here is the part almost nobody says out loud: an ATS does not reject anyone. It stores, sorts and searches. The rejecting is done by a person who never saw you.
What an applicant tracking system actually does
An ATS is, for the purposes of this page, a database with a search box on top. It keeps applications, pulls the text out of your file, sorts it into fields and lets whoever is hiring search and rank. Four things happen to your resume when it arrives, and all four can break.
The file is stored as it is
Your PDF goes in whole and stays there. None of this destroys your resume: the original is intact. What decides whether anyone ever opens it is the three steps below.
The text is extracted
An extractor walks the file and pulls out the strings of text with their position on the page. If your resume is an image, this step returns nothing: a picture of text contains no text, and the journey ends there.
The text is sorted into fields
The system tries to fill name, email, phone, employers, titles and dates by matching headings it knows. A heading you invented, like «My journey», is not one of them, and that section stays unclassified.
Everything is indexed for search
All that text becomes searchable. When a recruiter types «Google Ads» into the box, the system returns whoever carries those two words. Not showing up in that search is the only real damage there is.
Which produces the most important correction on this page: an ATS does not rule you out, because it has no judgement. A person rules you out, choosing among whatever that search box put in front of them. Your job is not to fool a machine. It is to not disappear before you reach anybody's eyes.
Where a resume breaks: the reading order
Here is the mechanism no guide explains and that decides nearly everything. A PDF does not store columns or tables: it stores fragments of text with coordinates. Rebuilding readable lines out of that is geometry, and the extractor solves it the simplest way: sweeping the page left to right and top to bottom.
Two columns
The sweep does not know there are two blocks, so it interleaves one line from each. «Excel March 2021» is your left-hand skill glued to a right-hand date. Your history turns to soup, and you get found for things you never wrote.
Tables
The same, with more noise. Plenty of templates put the dates in one cell and the job title in another, and on extraction the title and its dates come apart. A three-year role can end up reading as three months.
Headers and footers
It is where most people put their phone and email, and exactly the zone many extractors handle separately or skip outright. Your contact details are the last thing you can afford to lose.
Text boxes and icons
A text box floats outside the flow, so it lands wherever the sweep drops it. And that neat envelope icon in front of your email does not say «email» anywhere: it is a drawing, so the field arrives with no label.
Notice what they have in common: none of them is a content problem. Your experience is there and it is written down, but it arrives scrambled at the one place someone was going to look. Which is why the first rule of an ATS format is not to add anything. It is to take the layout away.
The format that does get read, point by point
All of the above is fixed with boring decisions. The full list, ordered by how much it hurts to skip:
One column, always
No exceptions. It is the only one here that fixes half the problems on its own, because it removes the interleaving. And it reads just as well on a screen.
A PDF whose text you can select
PDF is fine as long as you can select the text with your mouse. Open it, drag the cursor across a line and see if it highlights. If not, it is an image and it does not exist for anyone.
The headings everybody uses
Experience, Education, Skills, Languages. No «My journey», no «What I bring». The classifier looks for ordinary words, and creativity here is only ever paid for.
One date format, ten times
Pick one, «Mar 2021 to present», and repeat it in every entry. Mixing «03/2021», «March 21» and «since 2021» forces the system to guess how long you were there, and it guesses badly.
Contact details in the body, not the header
Name, email, phone and city in the first lines of ordinary text, one after another. It is where the extractor looks first and where they do not get lost.
A standard font, black on white
Arial, Helvetica, Calibri, Georgia. Decorative fonts embed badly and sometimes extract as broken characters. Nobody reads the colour: what gets read is the text.
A file name a human can parse
«Name-Surname-Role.pdf», not «resume_final_v3_good.pdf». It changes nothing technically, but it is the first thing a person sees in the list, and by then no machine is involved.
The demo shows a resume built from a concrete job ad, with the same line before and after. No signup.
Open the demoThe photo, and what it actually costs you
Whether a photo belongs on a resume depends on where you are applying, and in English-speaking markets it does not. Either way, it is worth knowing what happens to it: it is an image, so the extractor ignores it. It costs you no points, because there are no points. What it does is three things, and none of them helps.
It eats space that is worth more
A photo takes three to five lines out of the top third of page one, which is the most valuable part of the document. That space is worth more filled with your headline and your two best numbers.
It pushes your contact details down
Almost every template with a photo puts it top right, which usually means laying the header out as two blocks. You already know what happens to two blocks.
It adds a bias you cannot control
Your age, your gender and your face enter the decision before your experience does. In several countries a resume with a photo is set aside as a matter of company policy, precisely to keep that out.
The practical rule: if the ad does not ask for one, leave it out. And if it is a customer-facing role where one is genuinely expected, put it in the normal flow of the document, without splitting the header in two.
You can walk the whole product on fictional profiles, with the resume it builds for each ad.
Try the demoThe ad's own words, and how far copying them goes
The other half is not layout, it is vocabulary. Whoever is hiring searches that database by typing words, and they type them the way the ad they published wrote them. If the posting says «Google Ads» ten times and your resume says «SEM campaigns», you are the same person speaking two languages.
The manual method takes ten minutes: take the text of the ad, underline the nouns that repeat (tools, methods, the job title) and check which appear literally in your resume. Usually between eight and fifteen. The ones missing and true go back inside a sentence that says something, not a loose list at the bottom.
And here is the limit, which matters more than the trick: a word that is not true survives the search and dies in the interview. A tool you have never touched loses you the process for dishonesty, which is worse than losing it on fit. Use the ad's words only where you already were that person.
This part is what BeSuix does with the ad you paste: it pulls up to fifteen keywords out of it, marks which ones your profile covers and which it does not, and when it writes the tailored CV it weaves those words into your experience lines where they are true. The output is an A4 PDF, one column, selectable text, in one of two templates.
Keywords in the ad
Your CV for this ad
Before
Management of paid media campaigns on Google and Meta Ads.
After
Own paid acquisition across Google Ads and Meta Ads (~€35k/mo) and have taken blended ROAS from 2.1x to 3.4x over a year while holding CAC flat.
The rule
SQL stays out because it is not in your profile. Nothing you have not confirmed goes into the CV.
If you want the resume tailored to each ad without fighting the layout, access is free during the private beta.
Request accessWhat no ATS resume checker can tell you
Plenty of tools promise to «run your resume through an ATS» and hand you a score. It is worth knowing what they do: none of them has access to any ATS. They extract the text with the same sort of library anybody would use, then apply rules of their own. Useful, and not what the label says.
We say the same about ourselves: BeSuix does not scan your CV against a real ATS, and it cannot promise you will pass a filter, because nobody can promise that. What it does is score that application against that ad, and produce something machine-readable by design: one column, real text, fonts that embed cleanly.
What you can check yourself, in five minutes and with no tool at all:
Paste your PDF into a plain text editor
Open it, select everything, copy it and paste it into Notepad. What you see there is roughly what the system sees. If it comes out interleaved, missing accents, or empty, you have your answer.
Search it for your five keywords
Using the PDF reader's own find box, type the five words the ad repeats most. If one is missing and it is true of you, that is this evening's fix.
Email it to yourself
Open it on your phone, on another computer, in your mail client's preview pane. A font that embeds badly looks wrong exactly there, which is where the person hiring will open it.
That is everything format can do, and it is worth being clear about what it buys. Format does not win interviews. It stops you losing them for a reason that had nothing to do with you. The rest is the same work as always: choosing the right ads, and telling each one your experience in its own terms.
Frequently asked questions
Does an ATS reject resumes automatically?
Almost never in the way people mean. A recruiter can set a knockout question on a form, and that is automatic. The system itself only sorts and searches: a badly laid out resume never surfaces in that search, so nobody gets the chance to rule it out.
Word or PDF for an ATS resume?
PDF, as long as the text is selectable. Modern systems read it fine and it looks the same everywhere, which Word does not. Send a .docx only when the form asks for one. Never a scan, a photo, or anything exported as an image.
Can I put a photo on an ATS resume?
The parser ignores it, so technically it breaks nothing unless the layout around it splits the header into two blocks. The real cost is the space and the bias. If the ad does not ask for it, that space is worth more filled with your work.
How many pages should an ATS resume be?
One or two. Length is not a filter for the machine, which reads ten pages as happily as one, but it is a filter for the person. Under ten years, one page. Over that, two, with the relevant part finished before page one ends.
Are online ATS resume checkers worth using?
As an extraction test, yes: if one returns your resume scrambled, that is real information. As a score, treat it carefully. None of them is connected to a real applicant tracking system, so the number is their own rules, not a verdict from the software that will receive you.
Can I use a Canva or a designer template?
Only if it is one column, with no text boxes, no tables and no side panel, which rules out most of the ones that look good in the preview. The test is the one above: export it, paste it into a plain editor, see whether it still reads in order.
Keep reading
A CV written by AI, and how to find what it made up
The signals that give away an AI-written CV, what a generic tool invents and why, and how to audit yours line by line before you send it.
The questions on the application form, where most people give up
The four kinds of open question a form asks, how a 20-word answer differs from a 200-word one, and what to do when there is a character limit.