Parse first, rank second: what actually happens to a CV inside an ATS

#ai #career #careerdevelopment #resume

TL;DR

An ATS runs two distinct stages. It parses your document into structured fields, then ranks those fields. Parsing failures masquerade as ranking failures.
Document headers, two-column layouts, layout tables and text boxes are the four reliable ways to lose data before scoring begins.
Modern ranking uses semantic similarity, so keyword repetition has sharply diminishing returns and can be penalised as spam.

I build CVPilot, which means I have read more applicant tracking system parser output than is good for anyone. The consistent finding is that candidates optimise the ranking stage while losing at the parsing stage, which happens first and is entirely mechanical. If the extractor never associates a job title with an employer and a date range, nothing downstream can fix it.

Around 98% of large employers now screen CVs with software before a human reads them. That figure gets quoted so often it has stopped meaning anything, so here is the part that matters: the software changed in the last two years, and most CV advice did not keep up.

The old game was keyword counting. Repeat "project management" enough times and you cleared the threshold. Modern systems apply semantic matching, which means they understand that "managing projects" and "project management" are the same thing. Stuffing no longer helps, and on several platforms it now actively lowers your score.

This guide covers what an applicant tracking system does to your CV in 2026, the format that survives parsing, how to mirror a job advert without stuffing it, and a checklist to run before you apply. It follows UK conventions throughout.

What an ATS actually does to your CV

An ATS performs two separate jobs, and candidates usually only worry about the second one. First it parses, then it ranks.

Parsing converts your document into plain text and tries to map fragments into structured fields: name, contact, employer, job title, dates, skills. Ranking then scores that structured data against the job description.

Almost every failure people blame on ranking is really a parsing failure. If the parser could not find your job titles, no ranking algorithm can rescue you. I have written separately about why ATS systems reject UK CVs, and the causes are almost always structural.

Key Takeaway: Parsing happens before ranking. Fix the structure before you touch the wording.

The format that survives

Reverse-chronological, single column, standard headings. That is the entire answer, and it has been the answer for a decade, which is why the market is flooded with prettier advice that does not work.

File type

If the advert names a format, use it. When it does not, Word (.docx) is the safest default because virtually every parser handles it accurately. PDF is not inherently risky, but the tool that produced the PDF is.

Format	Parse reliability	Use when	Risk
.docx	Highest	No format specified	Minor layout shift across Word versions
.pdf exported from Word	High	PDF requested	Headers still skipped
.pdf from a design tool	Low	Never	Text may not extract at all
.txt	Perfect	Pasting into a form	All formatting lost
.pages / .odt	Unreliable	Never	Often rejected on upload

What breaks the parser

Contact details in the document header. Many parsers skip headers and footers entirely, so you arrive anonymous.
Two columns. The parser reads across the page, splicing both columns into one scrambled line.
Tables used for layout. Dates detach from the roles they belong to.
Text boxes and graphics. Invisible. Whatever sits inside them does not exist.
Skill bars and rating dots. They convey nothing to software and nothing verifiable to a human.
Non-standard headings. "Where I have made an impact" does not map to a field. "Work Experience" does.

Key Takeaway: Every visual flourish is a parsing risk. A CV that looks plain on screen is a CV that arrives intact.

Section order that maps cleanly

Use the headings the software expects, in the order a recruiter scans them. Deviating buys you nothing and costs you fields.

Name and contact details, in the body of the document, first three lines.
Personal statement, three or four lines, loaded with the advert's language.
Key skills, a plain list, no bars or ratings.
Work experience, reverse-chronological, one role per block.
Education.
Certifications, if relevant to the role.

Write each role as Job Title, Employer, Location, Month Year to Month Year. Consistent date formats matter more than people expect, because inconsistent ones cause parsers to mis-assign employment periods and silently invent gaps in your history.

Key Takeaway: Boring headings and consistent dates are worth more than any design decision you will make.

Mirroring the advert without stuffing it

Modern ranking uses semantic similarity, so repetition has sharply diminishing returns. What still matters is using the employer's exact vocabulary for skills, tools and job titles.

If the advert says "management accounts" and your CV says "monthly reporting", you have described identical work in language nobody told the system to look for. Match their term once, in a real sentence, then move on. I go deeper on this in the ATS keywords that actually matter.

Here is the contrarian part. The strongest CVs contain surprisingly few keyword repetitions, because they spend their words on evidence. Recruiters in 2026 are screening for measurable outcomes, not vocabulary.

Before and after

Weak bullet	Why it fails	Rewritten
"Responsible for managing the team"	States duty, not result. No number.	"Led a team of 9, cutting ticket backlog 42% in two quarters"
"Excellent communication skills"	Unverifiable and universal	"Presented quarterly forecasts to the board, adopted as the reporting standard"
"Used Excel, Excel modelling, advanced Excel"	Stuffing. Flagged as spam.	"Built the Excel model behind a £2.4m procurement decision"
"Helped improve processes"	Vague ownership, no scale	"Redesigned onboarding, reducing time-to-productivity from 6 weeks to 3"

Key Takeaway: Borrow their exact terms once each. Spend everything else on numbers.

UK conventions that US templates get wrong

Most CV templates online are American, and copying them quietly costs you. Do not include a photograph, your date of birth, your marital status, your nationality or your full home address. Your city and a contact email are sufficient.

These omissions are not squeamishness. UK employers are wary of receiving protected characteristics they did not ask for, and several ATS platforms strip such fields anyway. Two pages is the norm. One page is fine early in a career. Three pages needs a reason, and academic or clinical roles are the usual one.

Use British spelling throughout: optimise, analyse, programme, organisation. A CV written in American English reads as recycled from a template, and recruiters do notice.

Key Takeaway: UK CVs are plainer, shorter and carry less personal data. Follow the local convention.

The pre-submission checklist

Run this before every application. It takes five minutes and catches most of what gets CVs binned.

Contact details in the body, not the header.
Single column. No tables, text boxes, graphics or skill bars.
.docx unless the advert says otherwise.
Standard headings: Work Experience, Education, Key Skills.
Consistent date format on every role.
The advert's exact terms for its top five requirements, once each.
A number in at least half your bullets.
British spelling throughout.
Two pages or fewer, unless the sector expects more.
Filename contains your name and the role.

If your CV and your cover letter disagree about your job titles or dates, the system flags the inconsistency long before a human sees either. Worth checking both together, and the ATS cover letter guide covers that side.

You can also score a CV against a specific advert rather than guessing. CVPilot parses it the way an ATS does and shows which must-have terms are genuinely missing, which is a faster feedback loop than sending forty applications and inferring the problem from silence.