We'd like to use Microsoft Clarity to see how visitors use this site. It records clicks, scrolling and page views. It runs only if you allow it. See our Privacy Policy.

Skip to content
Country:Global
Login

Inside the ATS: what happens in the six seconds after you hit apply

Submit, parse, rank, search, scan. The ATS pipeline step by step, including the new AI grading, and where good candidates fall out.

The ApplyHustle teamEditorialUpdated 9 min read
A hand holding a smartphone with a stopwatch reading zero.

You spend forty minutes tailoring an application, press submit, and the posting goes quiet. The silence feels like a verdict from a black box. It is not. Once you understand how applicant tracking systems work, the box turns out to contain a fairly ordinary piece of database software, and ordinary software is something you can prepare for.

Every mainstream applicant tracking system runs the same pipeline. Your file is parsed into structured fields, stored as a candidate record, surfaced to recruiters through search, filters and rankings, and finally reviewed by a human who spends only seconds on the first pass. Workday, Greenhouse, iCIMS, Lever and Taleo wrap that pipeline in very different interfaces, but underneath they run the same steps in the same order. Here is what happens, roughly second by second, after you hit apply.

  • Submit: your form answers and your file arrive together, and any answer the employer marked as disqualifying can close the application here.
  • Parse: the file becomes a record of fields, and the original is kept as an attachment.
  • Store: the record joins every other applicant for the role and stays searchable for future roles.
  • Rank: where the employer has switched it on, an AI feature grades or sorts the record against the job's qualifications.
  • Retrieve: a recruiter filters, searches and opens records.
  • Scan: a person decides keep or discard in seconds.

Second one: your file becomes a record

Ask what is an ATS and most answers describe a gatekeeper. A better description is a filing system. An applicant tracking system is the database of record for a company's hiring: every application, note, interview score and rejection lives inside it. The moment your file arrives, the software's first job is resume parsing, which means extracting your name, contact details, job titles, employers, dates, education and skills, and writing each one into its own database field.

The scale of adoption is why this matters. Jobscan's 2026 count found an applicant tracking system behind the careers pages of 487 of the Fortune 500, with Workday the most common at just over two in five. Outside the largest companies the numbers are less well measured, but if you apply through a company careers page, some version of this parsing step is very likely happening to your file.

The output of that step is a candidate record. Your original PDF is kept as an attachment, but the record is what the system works with from this point forward. It is what gets searched, filtered, ranked and displayed. A careers page upload, a job board integration and a one-click apply button all end in a record of this kind, though some integrations pass your profile fields across instead of parsing a file.

That is the single most useful fact about the whole pipeline. You did not really submit a document. You submitted raw material for a database row, and everything downstream depends on how cleanly that conversion goes.

What the parser keeps, and what it throws away

The parser keeps text it can confidently map to a field. It throws away nearly everything else, which includes almost all of your design decisions. Fonts, colours, icons, column layouts and graphics do not survive the trip.

Some of what it discards matters far more than aesthetics. Columns can interleave, tables can be read in the wrong order, text inside images disappears, and some parsers skip headers and footers, which is a serious problem if that is where your email address lives. The full list, with a worked example of a broken record, is in what an ATS actually rejects.

Parsing has improved. Modern systems use machine learning rather than crude pattern matching, and a plain, well-structured file usually comes through intact. But ambiguity still loses data. A section heading the parser does not recognise may see its contents dumped into a miscellaneous field, or dropped altogether. Standard headings, a single column and contact details in the body of the page are not stylistic preferences. They are instructions the parser can actually follow.

File format matters less than people assume, with one caveat. A simply laid out PDF and a .docx both parse well in modern systems, and the old advice to avoid PDF altogether is out of date. What no format survives is complicated structure, because the parser reads content, not containers. A scrambled table scrambles identically in either format.

Ranking and filtering: how applicant tracking systems build a shortlist

Once parsed, your record sits in a database alongside every other applicant for the role, often hundreds of them. No recruiter opens each one in turn. They query the pile, which is why understanding how applicant tracking systems work is really about understanding retrieval.

The first cut is usually filters: answers to application questions, location, and any hard requirements the posting set. The second is ranking, where the employer has switched it on. Workday's HiredScore grades applicants from A to D against the requisition's qualifications, Greenhouse's Talent Matching sorts them into Strong, Good, Partial and Limited matches, and Lever, iCIMS and Oracle offer their own versions. The third is search, meaning keyword queries and Boolean strings run against the parsed text of every record.

None of this is judgement in any human sense. It is retrieval and sorting, closer to a library catalogue than to an interview. Records that contain the queried terms, or that match the qualifications the ranking checks, surface; records that do not stay buried. That is why recruiters search resumes rather than read them, and why the vocabulary on your resume matters as much as the experience it describes.

The specifics of the filters shift by country. A posting in the United States may filter on work authorisation, a British one on right to work, others on visa sponsorship or notice period. The mechanism is identical everywhere. Recruiters also save these searches and reuse them across roles, so a record that matches well for one posting keeps surfacing for the next.

The recruiter's screen: what they actually see

When a recruiter opens the role, they are not looking at a stack of PDFs. They see a list of parsed records: name, current title, current employer, dates, perhaps a snippet of matched text. Clicking a candidate usually opens the structured profile view first, your history rendered in the system's own plain formatting, alongside tags and notes left by colleagues.

Search results often highlight the matched terms in a snippet, the way a web search does, which quietly rewards resumes whose keywords sit inside meaningful sentences. Records also carry a status, advanced, rejected, on hold, and changing it takes a single click. That click, not any algorithm, is the moment most applications actually end.

A schematic recruiter screen showing a list of parsed candidate rows on the left and one expanded structured profile on the right with the searched terms underlined in a matched snippet, next to single-click advance and reject controls, illustrating that screening happens on the stripped-down record rather than the designed document.
Screening happens on this stripped view of the parsed record, and the status change behind the advance or reject chip is the click where most applications actually end.

ATS resume screening happens on this stripped-down view. The designed document you agonised over may never be opened at all. Recruiters can view the original attachment, and many do at later stages, but during screening the parsed record is the resume as far as they are concerned.

The practical consequence is blunt. Your first impression is not your typography. It is whether the record is complete, whether the titles and dates parsed cleanly, and whether the words the recruiter searched for are actually present.

Six seconds of human attention

Eventually a human reads. The six-second figure comes from eye-tracking studies run by the job site TheLadders, which measured about six seconds in 2012 and 7.4 seconds in a 2018 repeat. Both were small and run by a company with an interest in the answer, so treat the number as an order of magnitude, but it matches what recruiters report about screening at volume.

What matters more than the number is the decision being made. A six-second scan is not choosing who to hire. It is choosing keep or discard, deciding whether your record stays in the pile that earns a proper read. The eyes go to current title, current employer, dates, previous role and education, in roughly that order. If the scan finds a plausible match, you win a slower second read. If it finds nothing, the next record is one keystroke away.

An annotated parsed record showing the order a screening recruiter's eye moves in six seconds: current title first, then current employer, dates, previous role and education, connected by a dashed gaze path, with a note that the scan decides keep or discard rather than who gets hired.
The first scan lands on title, employer and dates in roughly that order, and the six second figure is a commonly cited order of magnitude rather than a precise measurement.

The arithmetic explains the brevity. A posting that draws three hundred applicants would take a full working day to screen at ninety seconds each, on top of every other role the recruiter carries. Nobody gets ninety seconds. This is also why the popular image of a machine coldly rejecting applicants is mostly wrong, a point unpacked properly in who actually rejected your resume. The filter with the highest kill rate is a busy person moving quickly through a stripped-down list.

Where good candidates fall out, and how to stay in

Put the pipeline together and the failure points name themselves. Most so-called ATS rejections are not rejections at all. They are parsing failures that mangled the record, keyword misses that kept it off the first page of search results, or a rushed human skipping a record that gave them nothing to hold onto in six seconds. Every one of those is preventable.

The fixes follow directly from the mechanism. Give the parser a single column, standard headings and contact details in the body of the page. Mirror the posting's vocabulary for the skills and tools you have. Put your strongest, most relevant material where a scanning eye lands first, at the top of the page and at the start of each role, and keep dates consistent so a stripped record shows no gaps a person would have asked about.

Then check your work at the level that counts, the record rather than the document. Read what an application form fills in from your file before you submit, and run the file through ApplyHustle's free ATS check, which flags a file it cannot read and shows which of a posting's keywords and requirements your resume covers.

The system is not your enemy. Its opacity is. Six seconds is plenty of time to be kept, provided the machine handed the human a clean record and the record leads with the right words.

How an ATS works: short answers

Does the recruiter see my PDF or the parsed version? Usually the parsed version first: a list of records, then a structured profile, with your original file one click away. Many recruiters open the PDF later in the process, but screening runs on the record.

How long does a recruiter look at a resume? The best-known measurement, from TheLadders, is about seven seconds for the first keep or discard decision. A resume that survives that scan gets a longer read.

Does AI decide who gets shortlisted? It decides the order in which a recruiter sees applicants, where the employer has turned it on. The shortlist itself is a status a person changes.

Share this article

Keep reading