Drafting a CV Without Getting Caught in the ATS Filter: How 2026 Algorithms Work

An applicant tracking system (ATS) is first of all a database. Employers use it to collect applications, move candidates through hiring stages and search across everyone who applied. Most of what candidates call "the ATS filter" happens in a few mechanical steps before any ranking or AI feature comes into play, and each step depends on the one before it. Once you know how ATS works step by step, most of the advice about CV layout explains itself.

Step 1: text extraction

When you upload a file, the system pulls out its text layer. Fonts, colours, lines and icons are dropped; only characters remain, in the order the file stores them. In a plain single-column document, that order matches the reading order. In a designed layout it often does not.

A two-column CV is the classic case. Many parsers read straight across the page, so a line from the sidebar and a line from the main column end up joined:

Your job title now sits under a skills heading, and your dates are glued to a skill list. There is more on this in two-column CVs and ATS.

Some files have no text layer at all. A CV exported as an image, or scanned, contains pictures of letters. Whether the system runs optical character recognition (OCR) on it depends on the product and its settings; if it does not, the page arrives empty. Skill bars, star ratings and icons cause the same problem on a smaller scale: a bar filled to 80 percent next to "Python" carries no text, so the skill level disappears.

Step 2: field mapping

Next, the parser tries to split the text into fields: name, email, phone, job titles, employers, dates, education and skills. It relies on cues: conventional section headings, recognisable date formats and a predictable order such as title, company, dates. Headings like "Experience", "Education" and "Skills" are easy to map. Creative ones such as "My journey" or "Toolbox" may not be recognised, and the content under them can land in the wrong field or in none.

Dates matter more than they look. A filter that asks for three years of experience compares a number the parser calculated from your date ranges. If a date is split across columns or written in an unusual format, that number can come out wrong.

Step 3: search, filters and ranking

Once your profile is stored, recruiters work with it. What happens next depends on how the employer set the system up:

• Knock-out questions in the application form, such as a work permit or a required licence, can screen candidates out automatically.
• Keyword search lets a recruiter find candidates whose text contains specific terms, often matched literally.
• Matching or ranking features, where they are switched on, compare the CV with the job ad and sort applicants. Newer systems add semantic matching that also counts related terms.

None of these can work with text that was lost or scrambled in steps 1 and 2. That is why readability comes before keyword optimisation.

Step 4: a person reads it

Shortlisted CVs are read by a recruiter or hiring manager, sometimes in the parsed view and sometimes as the original file. A CV that survived parsing still has to make sense to a person in a short read.

What this means for your CV

• Use a single column for the main content. Keep tables and text boxes out of experience and skills.
• Export a real text PDF or a DOCX from your word processor; do not scan it or flatten it into an image. If you are unsure which format to send, read PDF or DOCX for your CV.
• Use standard headings and one date format throughout.
• Write skill levels as words ("Python – advanced"), not bars.
• Put contact details in the body of the document, not in the header or footer.

To see your own CV the way a parser does, the parser view lists every extracted field and flags the ones that look split, suspect or missing. For a fast first check of structure, you can paste your text into the quick ATS test.