What an applicant tracking system actually does

A database with a workflow attached

An applicant tracking system (Workday, Greenhouse, Lever, Ashby, Taleo, SuccessFactors, iCIMS, Zoho Recruit) is a database of applications with a workflow on top. It does four things:

1. Parses your CV into structured fields: name, contact, employers, dates, titles, education, a bag of skill keywords.
2. Stores it, permanently, along with every previous application you made to that company.
3. Lets recruiters search and filter: by keyword, by location, by answers to the form’s questions.
4. Tracks stage and sends templated email.

That is the whole machine. It is unglamorous and it is not the villain of the story.

The score is a product, not a feature

You will be offered tools that “scan your CV like an ATS” and return a match percentage, with a subscription to improve it. No mainstream applicant tracking system assigns your CV a percentage and rejects you below a threshold. The number is invented by the company selling you the report.

What is real, and different in an important way:

• Recruiters can sort and filter. If a recruiter searches the pile for “dbt” and your CV never contains the word, you are not in the results. That is a human choosing a filter, not an automatic rejection, but the effect on you is the same, which is why the words matter.
• Knockout questions are automatic. The form asks whether you have the right to work in the country, whether you have five years of experience, whether you can be on site. Answer in the disqualifying direction and you are rejected without a human involved. This is where automatic rejection actually happens, and it happens on the form, not in your CV.
• Some systems rank candidates and surface a suggested order. Recruiters vary in how much they trust it, and regulators are increasingly interested in it: New York City requires bias audits of automated employment decision tools, and the EU AI Act treats recruitment systems as high risk. The rules here are still moving and differ by jurisdiction.

What parsing actually breaks on

Since the machine’s real job is reading your document into fields, the practical advice is about being readable:

• One column. Two-column layouts frequently parse as interleaved nonsense: your job titles woven through your skills list.
• No text in headers or footers. Many parsers drop them entirely, which is how people submit a CV with no phone number.
• No tables for layout, no text inside images, no icons standing in for words. A skills chart of five-star ratings parses as nothing at all.
• Standard headings. “Experience”, “Education”, “Skills”. Not “My Journey”.
• Dates in a consistent format, ideally Mar 2023 – Jun 2025.
• The file type the form asks for. If it says PDF, send PDF; if it says .docx, send .docx.

You can check yours in thirty seconds, free, without any tool: open your PDF, select all, copy, and paste into a plain text editor. What you see is roughly what the parser sees. If your job titles are scattered and your contact details are missing, fix the layout: that test is worth more than any paid report.

The words

Use the advert’s exact terms where they are true of you. If they say “ETL” and you wrote “data pipelines”, write both once. If they say “stakeholder management” and you did it, use their phrase.

What does not work: keyword stuffing, white text on white background, or a hidden block of terms. These get you rejected by the human two minutes later, and some systems flag them. The invisible-text trick in particular is recirculated every year and it has never worked.

The two things that matter more than any of this

Apply through the company’s own careers page where one exists. Applications through aggregators sometimes arrive as an email attachment nobody opens, or never arrive at all.

Get a human to look. A referral, an email to the hiring manager, a comment from someone internal: any of these routes around the pile entirely, and that is the actual answer to “my CV is not getting through”.

The limitation

The formatting advice above is worth doing once, in an hour, and then never thinking about again. Parsing is not what is stopping you. Nine hundred other applications are. Treat the ATS as a door you should not accidentally lock rather than as the reason it will not open.