What an ATS actually reads in your CV

Most advice about applicant tracking systems is written by people who have never seen one reject a file. The result is a folklore of rules (never use a table, always use Arial, keyword-stuff the skills section) that is partly true, partly superstition, and rarely explained.

Here is the part that is worth knowing: an ATS does not look at your CV. It looks at the text it managed to extract from your CV. Everything else follows from that.

The parse is the whole game

When you upload a PDF, the system runs it through a text extractor. What comes out is a stream of characters in whatever order the extractor found them. That stream is what gets stored, searched, and scored. Your careful two-column layout, your section dividers, your icons: none of it exists at that point.

This is why the common advice half-works. A table is not forbidden; it is risky, because some extractors read a two-column table row by row and interleave the columns:

Senior Engineer 2021–2024 Acme Ltd Built the billing system Junior Engineer 2019–2021

That is still readable to you. To a keyword matcher looking for "Senior Engineer" followed by a company, it is noise.

What survives reliably

• Plain paragraphs and single-column lists. These extract in the order you wrote them, every time.
• Real text. If you can select it in a PDF viewer, the extractor can read it.
• Standard section headings. "Experience", "Education", "Skills". Not because the software is impressed by them, but because many parsers use them as anchors to split the document into fields.
• Dates in a consistent format. 2021–2024 or Jan 2021 – Mar 2024. Mixing formats inside one CV makes the date parser guess.

What gets dropped or mangled

• Text inside images. A logo with your name in it is invisible. So is a "skills wheel" graphic.
• Headers and footers. Many extractors skip them entirely. If your phone number and email live in the header, the system may have no way to contact you, a failure that looks exactly like being rejected.
• Text boxes and columns, depending on the extractor. This is the one that varies most between systems, which is why the advice is inconsistent.
• Unusual bullet glyphs. Rare, but a decorative character can end up glued to the first word.

The two-minute check anyone can do

Open your CV in a PDF viewer. Select all the text, copy it, and paste it into a plain text editor.

What you see is approximately what the ATS sees.

If your job titles are interleaved with dates, if your contact details are missing, if entire sections came through as nothing, that is not a formatting preference, that is your application arriving incomplete. Fix that first, before you think about keywords.

About keywords

Keyword matching is real, but it is cruder than people assume. Most systems are doing something close to: does this document contain the terms from the job description, and how often.

Two practical consequences:

1. Use the words the posting uses. If it says "stakeholder management" and your CV says "worked with clients", a literal matcher scores you lower, not because you lack the skill, but because it is not looking for the skill, it is looking for the string.
2. Do not stuff. Repeating a term twenty times in white text is an old trick that modern systems flag and humans find insulting when they open the file.

The honest version of "optimise for keywords" is: describe what you genuinely did, using the vocabulary the employer used. That is a rewriting job, not an invention job.

Where the line is

There is a real difference between tailoring and lying, and it is worth being clear about it because the tools in this space blur it.

Tailoring is choosing which of your true experiences to lead with, and describing them in the language of the role, which how to tailor your CV to a job posting works through step by step. If you built an internal reporting tool and the posting asks for "data visualisation", saying so is accurate.

Inventing is claiming a technology you have not used, a title you did not hold, or a result you did not produce. It is also, practically, a bad bet: it survives the parse and fails the interview.

This is the rule RezoCV is built on: it rewrites from the evidence in your CV and will flag a gap rather than fill it with something plausible. A tool that quietly invents is not saving you work; it is writing a claim you will have to defend in a room.

What to do next

1. Do the copy-paste test above. Fix anything that comes through broken.
2. Read the posting and note the terms it repeats. Those are the ones being matched.
3. Rewrite your bullets in that vocabulary, keeping every claim true.
4. Export as PDF unless the form asks for DOCX, and if it is an ATS form, DOCX parses more reliably.

None of this is a trick. It is making sure the document you spent hours on arrives intact, and says what you did in words the reader is looking for.

The parse is one of five separate things an application has to get right. The five jobs of a job application covers the other four, and how to tell which one is costing you replies.