What an ATS actually does with your CV
An applicant tracking system is a database with a hiring workflow attached. When you submit an application it converts your file to plain text, attempts to split that text into fields — name, contact details, employment periods, titles, employers, education, skills — and stores the result as a record. The recruiter then searches and filters those records. This matters more than any styling question, because what the recruiter sees in their shortlist is the extracted record, not the document you were so careful about. Your PDF is usually still attached, but it is opened after a decision to look closer has already been made. The attachment matters less than people assume: a recruiter working a long shortlist filters and scans the records first, and opens files only for the handful that survive that pass.
Nothing in that sequence is an automated rejection. The system does not read your CV, form a judgement and decline you; it files you, and a human decides whom to look at using filters and searches over what got filed. The practical consequence is the same in one specific case, though: if the extraction lost the field a recruiter filters on, you are absent from the result set that decides who gets read. Not rejected — simply not present, which is worse, because nobody knows to look for you. So the useful question is never whether an ATS liked your CV. It is whether the record built from your CV still contains the things you wanted an employer to be able to search for.
What a checker can genuinely test
The distinction worth holding on to is between things that are observable from your file and things that depend on an employer's private configuration. A checker sees your document and, if you supply one, the job advertisement. It does not see the employer's ATS, their filters, their scoring rules, or the other applicants. Any claim that requires one of those is a guess dressed as a measurement, and you should treat a tool that makes such claims with suspicion — including when the tool is ours. The practical test is whether a finding can be checked by you independently. Anything phrased as a reading of your own document can be verified against the document; anything phrased as a prediction about an employer cannot.
| Claim a checker might make | Testable? | How, or why not |
|---|---|---|
| Your file converts to clean text | Yes | Run the same extraction and show the output |
| Your sections were recognised | Yes | Check which headings mapped to known fields |
| Your dates parsed correctly | Yes | Compare extracted periods against the document |
| Terms in the job ad are missing from your CV | Yes | Direct comparison of two texts |
| Your CV will pass this employer's filter | No | Their filter settings are private |
| Your CV scores 78 out of 100 | No | No shared scale exists; the number is the tool's own |
| You were auto-rejected by an ATS | No | Mainstream systems do not auto-reject |
| You rank above other applicants | No | The tool cannot see the other applicants |
The parsing failures that actually cost interviews
In practice a small number of layout decisions cause most extraction damage, and they are all invisible when you look at your own CV, because your eye reconstructs what the parser cannot. Two-column layouts are the worst offender: the parser reads in text order, which on a two-column page means it interleaves the columns and produces a record where your job titles have your skills spliced through them. Text inside an image or a graphic extracts as nothing at all, so a designed header containing your name and contact details can produce a record with no name in it.
Dates are the most frequently corrupted field. A parser is looking for a start and an end, and it copes badly with mixed formats, seasons, ranges written with en-dashes in some rows and hyphens in others, or a present role marked with a word rather than a date. Pick one format, use it in every row, and prefer numerals. The second most common loss is anything placed in the page header or footer region, which a number of parsers skip entirely — a phone number that lives only in the footer may not exist as far as the record is concerned. None of this is visible to you afterwards, which is the reason it persists: the document on your screen is unchanged and perfectly legible, and the damaged copy exists only inside a system you cannot see.
Keyword matching, and how far it really goes
Keyword matching is real. Recruiters do search their database for the terms in the requisition, and a CV that never uses the employer's vocabulary is harder to find. This is the legitimate core of what checkers do: compare the language of the advertisement against the language of your CV and show you what is absent. Where it gets overstated is the leap from that to a percentage of fit, as though hiring were a text-similarity problem. Matching tells you whether you are findable, not whether you are suitable, and the two get conflated constantly in marketing copy.
The honest use of a match report is as a prompt, not an instruction. If the advertisement asks for something you have genuinely done and your CV does not mention it, that is a real gap and worth fixing in your own words. If it asks for something you have not done, the report has found a mismatch between you and the role, and inserting the term anyway is how people end up in interviews they cannot survive. A tool that encourages you to paste in missing keywords regardless is optimising the metric rather than your outcome.
See what a parser actually extracts from your CV
AppCV runs the extraction and shows you the result — which sections were recognised, which dates survived, and which terms from a specific job advertisement your CV never uses. The check is free and needs no card.
The myths that survive because nobody checks them
- Myth: the ATS scores you and rejects you. It files you; a recruiter decides.
- Myth: you must hit an exact keyword density. There is no such threshold in any mainstream system.
- Myth: PDFs are unreadable. Text-based PDFs parse fine; scanned or image-based ones do not.
- Myth: white text hides keywords for the parser. It is visible in the extracted text and reads as deception.
- Myth: one universal template beats every ATS. Systems differ; the shared requirement is plain structure.
The formatting rules that do hold up
Strip away the folklore and a short, dull list remains, which is good news because it is easy to comply with and costs you nothing stylistically. One column. Real text rather than images. Conventional section headings, because parsers match on the words they have seen a million times rather than on your phrasing. One date format throughout. Nothing load-bearing in the header or footer. Standard fonts, because an exotic embedded font can extract as wrong characters. None of that requires an ugly document; it requires a structurally simple one.
- Export as a text-based PDF unless the employer explicitly asks for Word.
- Use one column for the whole document, including the contact block.
- Name your sections conventionally: Experience, Education, Skills.
- Write every period as MM/YYYY, in every row, with no exceptions.
- Keep your name, phone and email in the body of the page, not the header region.
Why a score without a job advertisement means little
A checker that gives you a number without asking which job you are applying for is scoring your formatting and calling it fit. Formatting is worth checking, but it is a pass-or-fail property rather than a graded one: either the parser read your document or it did not. The interesting question is always relative to a specific posting, because the same CV is strong for one role and irrelevant for the next. Treat an absolute score as a rough hygiene signal at best, and be sceptical of any tool that displays one prominently — a single number is easier to sell than a useful report.
Free and paid checkers
Most checkers are free at the point where they tell you a problem exists and paid at the point where they help you fix it. That is a reasonable model, and it is roughly ours, but it creates an incentive worth naming: a tool earns more when it finds more problems, which is a quiet pressure toward inflating the list. The tell is whether the findings are specific and checkable — "your 2019 date parsed as 2010" is a fact you can verify, while "your CV needs optimisation" is a sales prompt. Prefer a tool that shows you its working.
What no checker can tell you
No third-party tool sees the employer's actual configuration: which system they run, which fields they filter, what their recruiter typed into the search box, or how many other people applied. Anyone claiming otherwise is guessing. Nor can a checker judge whether your experience is competitive, whether your career change reads as coherent, or whether the person reading will like you. Employers themselves report that their hiring filters screen out qualified candidates — 88% of them, in a survey covering Germany — which tells you the problem is real, and also that no external tool can see inside it.
That is the honest boundary of this category, and it applies to us as much as to anyone. A check tells you whether your document survives being read by software and whether it speaks the language of a specific advertisement. Those are worth knowing and they are cheap to fix. They are not the same as being the strongest applicant, and a high score is not an interview. Any tool that lets you believe otherwise is selling you reassurance rather than information.
Run the check against a real job advertisement
Paste the posting you are actually applying for and AppCV scores your CV against it, lists the terms it asks for that your CV never mentions, and shows what the parser extracted. It cannot see the employer's own filters, and it will not tell you whether you are the strongest candidate — no tool can.
Frequently asked questions
- What is an ATS resume checker?
- It is a tool that runs your CV through the same parsing step an applicant tracking system performs, then reports what survived: which sections were recognised, whether your dates and job titles extracted correctly, and — if you supply a job advertisement — which of its terms your CV never uses. The useful ones score against a specific posting. A checker that returns a number without asking which job you are applying for is grading your formatting and presenting it as fit.
- Are ATS checkers accurate?
- They are accurate about what they can observe and speculative about the rest. Whether your file converts to clean text, whether your sections and dates parsed, and which terms from a job ad are missing are all directly checkable. Whether you will pass a particular employer's filter is not: their configuration is private. Treat extraction findings as facts you can verify yourself, and treat any absolute score as one tool's opinion on its own invented scale.
- Do companies really use ATS?
- Yes, for almost any application submitted through a web form. An applicant tracking system is the standard way employers receive, store and search applications, and large employers rarely operate without one. What is exaggerated is what it does: it files and indexes your application so a recruiter can search it. Mainstream systems do not read your CV and reject you automatically, and the widely quoted claim that they discard most CVs has no checkable source behind it.
- What ATS score is good?
- There is no industry scale, so the question has no general answer — a 90 from one tool and a 60 from another can describe the same CV. Scores are only meaningful relative to a specific job advertisement and only within one tool, for comparing a before and an after. Use a score to see whether an edit improved things, never as a threshold to clear. The parsing findings underneath the number are the part that actually tells you what to change.
Related guides
Published 21 September 2026