What Is an Applicant Tracking System and How Does It Work

Almost everyone applying for jobs today has heard that an ATS will read their CV before a human does. Far fewer people could say what one is, and the gap between those two facts is where most bad CV advice lives.

An applicant tracking system is a database. That is the honest one line answer, and it deflates most of the folklore attached to the term. It is the software a company uses to collect applications, store them, search them, move candidates between stages, and keep a record of who was contacted and when. Workday, Greenhouse, Lever, SuccessFactors, Taleo, iCIMS and a dozen others are all doing roughly this job.

It is not a robot that scores you out of a hundred and bins you at 74. That version exists mostly in the marketing of companies selling you a fix for it.

But the database is real, and how it stores your document has real consequences for whether a person ever reads it. That is worth understanding properly.

What actually happens after you click apply

Follow a single application through the system and the mystery mostly evaporates.

You upload a file. Usually PDF or Word, occasionally through a portal that asks you to paste text or connect a LinkedIn profile.

The system parses it. Parsing means converting your visually formatted document into structured data fields: first name, last name, email, phone, then a list of jobs, each with an employer, a title, a start date and an end date, then education, then a bag of skills. Your CV goes in as a layout and comes out as rows in a table.

You get a form to correct. Many systems then show you the parsed result and ask you to fix it. That screen is the single most useful piece of feedback in the whole process, and most candidates click through it in irritation without reading it. If the form has your job titles scrambled or your dates missing, that is not a form bug. That is what the employer's copy of your CV now looks like.

The record lands in a pipeline. A recruiter sees a list of candidates for the role, with columns drawn from the parsed fields, and a status such as New, Screened, Interview or Rejected.

Someone searches or filters. This is the step people imagine as the machine judging them. It is closer to a recruiter typing a search term into a box.

A human opens your CV. In almost every case, the original file is right there, one click away, and a recruiter who is interested will open it.

The important detail in that sequence is that there are two versions of you inside the system: the structured record built by the parser, and the original document. Search and filtering run on the structured record. Judgement runs on the document. Your job is to make sure the first one is accurate enough to get someone to open the second one.

What the system genuinely does

It parses. Reliably for plain, linear layouts. Less reliably for anything clever. Two column designs are the classic failure, since the parser may read straight across the page and interleave two unrelated columns into nonsense. Text inside images cannot be read at all. Content in the header or footer area of a Word document is frequently dropped.

It searches. A recruiter with 400 applications for one role does not read 400 CVs. They search for a term, look at the 30 results, and start there. If your CV never contains the term they searched, you are not in that 30, and nothing about your quality changes that.

It filters on knockout questions. Those questions during the application about work authorisation, notice period, salary expectation, years of experience or a required licence are the real automated gate. Answer one in a disqualifying way and you can genuinely be filtered out without a human involved. This is far more consequential than anything to do with formatting, and it is the part almost nobody worries about.

It deduplicates and remembers. Apply to the same company three times and the system links those applications. Recruiters can see your history, including previous rejections and any notes attached to them.

It tracks compliance. A large part of why these systems exist at all is the legal requirement to show a consistent, documented hiring process. That shapes the software much more than any candidate ranking ambition.

What it mostly does not do

It does not reject you automatically because of a low score. Some systems offer a match percentage. It is a sorting aid shown next to a candidate, not a gate, and recruiters routinely ignore it. There is no industry standard 75% cutoff, whatever a template site tells you.

It does not read your cover letter and grade your enthusiasm. Usually the cover letter is stored as an attachment and opened only if someone is already interested.

It does not punish you for using the word "the" too often, or for having a two page CV, or for a serif font. These are invented rules. Page count is a human convention and the parser is entirely indifferent to it, as covered in how long a CV should be.

It does not detect white text keyword stuffing and blacklist you. The parser will happily read invisible text. A human then opens the document, sees a wall of hidden keywords when they select the text or convert it, and forms a rather firm view of your honesty. The mechanism that punishes this trick is embarrassment, not software.

The keyword question, answered honestly

Keywords matter, but not for the reason usually given. They matter because of search, not scoring.

Picture the recruiter. They have a requisition for a data analyst. They open the pipeline, and rather than reading everything they search for "SQL", or "Power BI", or whatever the hiring manager stressed in the briefing. The system returns everyone whose record contains that string. That is the whole mechanism.

Three consequences follow.

Use the employer's words, not your preferred synonyms. If the advert says "customer success" and your CV says "client relationship management", a search for the first term does not find you. You are not being scored down, you are simply absent from the result set. This is the practical core of tailoring your CV to a job description.

Spell out abbreviations once. Write "search engine optimisation (SEO)" the first time. Different recruiters search for different forms of the same thing, and one line covers both.

Put the term where a human can also see it working. A keyword sitting in a skills list proves nothing. The same keyword inside a bullet that shows what you did with it survives the human read that follows the search. Guidance on getting that balance right is in what to include in a CV skills section.

Keyword stuffing fails on exactly this point. It gets you into the search results and then loses you at the moment a person actually reads.

Where applications genuinely die

Having established what the software does, here is the more useful ranking of what actually removes candidates, roughly in order of how often it happens.

A human read it and was not convinced. This is the overwhelming majority of rejections. The CV parsed perfectly, the recruiter opened it, and the experience did not match what they needed. No technology was involved. It is worth sitting with this, because it is much easier to blame a machine than to accept the document needs to be better.

A knockout question was answered in a disqualifying way. Genuinely automatic, genuinely common, and entirely under your control if you read the questions carefully.

Nobody ever searched a term your CV contains. Fixable by using the employer's vocabulary.

The parse was so broken the record was unusable. Real, but rarer than the internet suggests, and almost always caused by a specific identifiable choice: a two column template, a photo containing text, contact details in a header, dates written in an unusual format, or job titles styled to look like section headings.

You applied to a role that was already effectively filled. More common than anyone likes to admit. Internal candidates and referrals fill a large share of advertised roles, and the posting stays open for process reasons.

Notice that only one of those five is a formatting problem, and it sits fourth.

What to actually do about it

The practical guidance is short, which is a good sign.

Use a single column layout with plain section headings: Experience, Education, Skills. Put contact details in the body of the document rather than in a header or footer. Export as PDF unless the posting explicitly asks for Word. Write dates in a consistent, unambiguous format, with the month named or numbered the same way throughout. Avoid tables, text boxes, icons carrying meaning, and images containing words. Use the vocabulary from the advert where it honestly describes your work.

Then stop thinking about the machine and spend your remaining effort on the part that decides the outcome: whether your bullets show results rather than duties, and whether the top third of the page makes the case for this specific role. The full checklist is in how to write an ATS friendly CV, and the most common self inflicted wounds are collected in common CV mistakes to avoid.

One more thing, and it may be the most valuable habit in this entire article: when the application form shows you the parsed version of your CV, read it. That screen is a free preview of what the employer sees. If your most recent role is missing or your dates are wrong, fix the document rather than patching the form, because the same document is going to the next twenty employers too.

The systems behave differently, and it is worth knowing how

Treating "the ATS" as one thing is part of what makes the advice around it so confused. The systems differ enough that the experience of applying tells you something about what you are dealing with.

Modern systems built in the last decade (Greenhouse, Lever, Ashby and similar) tend to have clean application forms, decent parsers, and very little in the way of automatic scoring. They are built around structured interview processes and scorecards filled in by humans. If the form was pleasant to use, you are probably in one of these, and formatting anxiety is largely wasted energy.

Enterprise suites (Workday, SuccessFactors, Taleo and their relatives) are modules bolted into a much larger human resources platform. They are the source of most complaints about applying: the account creation, the twenty minute form, the request that you retype the CV you just uploaded. Their parsers are older and less forgiving, and the retyping request is exactly why. When one of these asks you to check the parsed fields, take it seriously.

Job board pipelines (LinkedIn Easy Apply, Indeed and similar) are a slightly different animal. Here the profile often matters more than the attachment, because the recruiter is filtering inside the board's own interface. That is the practical reason to keep a profile consistent with your CV, which is the subject of optimising your LinkedIn profile for recruiters.

You cannot tell in advance which one a given employer runs, which is a good argument for a CV that is plain enough to survive the oldest parser you might meet.

What happens to your application afterwards

One genuinely useful thing about a database is that it remembers, and candidates rarely think about what that means.

Your record does not disappear when you are rejected. It stays in the system, tagged with the role you applied for, the stage you reached and any notes a recruiter wrote. Recruiters at that company can search the existing database before advertising a new role, which is why people sometimes get contacted about a job they never applied to. A near miss for one role is a genuine asset for the next one at the same employer.

It also means inconsistency is visible. If you apply twice in six months with different job titles or different dates for the same period, both versions sit in the same record. Nobody is running a fraud investigation, but it looks careless, and careless is not the impression you want attached to your name in a database.

Under European data protection rules you can ask what a company holds about you and ask for it to be deleted, and most systems have a retention period after which records are purged automatically. If you would rather not stay in a company's talent pool, that request is yours to make.

Frequently asked questions

Do all companies use an ATS?

Most medium and large ones do, and many small ones use a lightweight version through a job board or recruitment agency. If you applied through a web form rather than by email, a system of some kind is holding your application.

Is a PDF safe for an ATS?

Yes. Text based PDFs have been parsed reliably for years, and PDF preserves your layout for the human reader. The exception is a PDF that is really a scan or an exported image, which contains no text at all and parses to nothing. The trade offs are covered in more depth in the file format guidance.

Does a two page CV hurt me with an ATS?

No. The parser reads the whole document. Page count is a convention for human readers, and two pages is standard for experienced candidates in most of Europe.

What is a good ATS score?

There is no such thing as an official one. Scores shown by CV checking tools are that tool's own estimate of how well your CV matches a job description. Useful as a mirror, meaningless as a grade, and not the number any employer sees.

Should I remove my photo because of the ATS?

The parser simply ignores it, so it is not a technical question. It is a regional convention: expected in Germany and much of Central Europe, avoided in the UK, Ireland and the US, where employers often prefer applications without one for anti bias reasons.

If a recruiter can just open my CV, does formatting matter at all?

Yes, for two reasons. A broken parse can keep you out of the search results that decide who gets opened, and the parsed fields fill the columns a recruiter scans before deciding to open anything. Good formatting does not win you the job, but bad formatting can stop the conversation before it starts.

See what a parser sees

The quickest way to stop guessing is to look at your own CV the way the software does. CvLaunch reads your file and shows you the extracted result: which roles, dates, sections and skills came through, and what got lost on the way. It takes a few seconds, it is free, and there is no sign up. If everything is there, you can stop worrying about the machine and get back to the part that actually decides it.