An applicant tracking system is mainly a database and workflow tool: it stores applications, moves them through stages and lets recruiters search. Most do not silently auto-reject on keywords. Real automated filtering usually comes from knockout questions you answered and from search ranking that decides who gets read first.
- The widely-quoted "75% rejected by ATS" figure has no traceable source. Treat it as folklore.
- The real filters are knockout questions and search ranking — not a secret keyword robot.
- Being unread is the common outcome, and it is different from being rejected.
- Parsing failures are real: complex layouts, tables, columns and images genuinely lose information.
- Optimise for a human skim first, machine parsing second, and never keyword-stuff.
First, the statistic
You have almost certainly seen a version of "75% of résumés are rejected by an ATS before a human sees them." Sometimes it is 70%, sometimes 88%.
There is no traceable primary source for it. It gets attributed to different vendors and consultancies depending on where you find it, and following the citation chain leads to other articles citing other articles. It should be treated as folklore, and it is worth naming because an entire industry of anxiety — and a lot of paid "ATS optimisation" — rests on it.
The honest version is less dramatic and more useful: many applications are never read. Not because a robot discarded them, but because a role received 600 applications and nobody read 600 applications. Being unread is a different problem from being rejected, and it has a different solution.
What an ATS actually is
Mostly a database with a workflow attached. Its core jobs:
- Store applications in one searchable place
- Parse the résumé into structured fields — name, employers, dates, education, skills
- Move candidates through stages, with notes and scores attached
- Search and rank so a recruiter can find relevant applications
- Communicate — templated emails, scheduling
- Report — funnel metrics, source tracking, compliance records
Notice what is not on that list: deciding. The archetypal ATS is an organiser, not a judge. Vendors do increasingly bolt on matching and scoring features, and some employers switch them on — which raises real questions, covered in what counts as an AEDT — but the base product is filing software.
Where filtering genuinely happens
Three places. None is a secret keyword robot.
1. Knockout questions. The most common cause of a genuine automated rejection, and the most overlooked. These are the questions on the application form: are you authorised to work in this country, do you require sponsorship, do you have X certification, will you relocate, what are your salary expectations. Employers configure some of these to auto-disqualify.
If you were rejected within minutes of applying, this is almost always why — not your résumé. Answer the questions carefully. A dropdown chosen carelessly is the single most preventable rejection in the process.
2. Search and ranking. A recruiter searches the applicant pool: "Kubernetes AND Terraform, applied in the last 14 days". You either appear in the results or you do not. If you appear on page four, you are functionally invisible.
This is where vocabulary matters. Not because a robot is scoring you, but because a human's query is made of words, and your document either contains them or does not. If the job description says "Kubernetes" and you wrote only "K8s", you may not appear in that search.
3. Parsing failures. Real, and the most fixable. If the parser cannot read your document, your structured profile is empty or wrong, so you do not appear in searches and your experience looks thinner than it is.
Known parser problems:
- Two-column layouts — content interleaves into nonsense
- Tables and text boxes — frequently dropped entirely
- Text inside images or graphics — invisible; an "infographic CV" can parse to almost nothing
- Headers and footers — sometimes skipped, so contact details vanish
- Non-standard section headings — "My Journey" instead of "Experience"
- Unusual date formats — breaks tenure calculation
- Fancy fonts and ligatures — can garble extracted text
What to actually do
Format for the parser, write for the human. Single column. Standard headings: Experience, Education,
Skills. Real text, not images. No tables or text boxes for anything important. Contact details in the body,
not the header. A common font. Consistent Month Year – Month Year dates. Save as .docx or a text-based
.pdf — and if you can select and copy the text in your PDF viewer, a parser can read it.
Mirror the employer's vocabulary for things you actually have. Their words for your real skills. This is matching, not stuffing.
Never keyword-stuff, and never list skills you lack. White text, hidden keyword blocks and skill lists that contradict the described experience all fail the moment a human reads the document — which they will, at exactly the stage that matters.
Answer the screening questions like they are the screen. Because they are.
Apply for roles you actually match, and apply early. Recency is a common sort order.
Do not pay for an "ATS score". Those tools do not have access to the employer's configuration and cannot know their search terms. They are checking your document against a generic heuristic and charging for the reassurance.
The part nobody says out loud
Everything above is advice on formatting a document so that software can extract data from it, so that a recruiter's search finds it, so that a human might read it and infer whether you can do the job.
That is four layers of proxy between you and the actual question — can this person do the work?
The résumé is a self-reported, unverifiable summary of the past, and it is the wrong artefact for the job. It rewards the people best at writing résumés, it carries identity signals that demonstrably change how the same content is read, and it needs an entire cottage industry to explain how to survive it.
The alternative is not a better-formatted résumé. It is not having the résumé be the artefact — see what skills-based hiring actually is and getting hired without a degree.
How it works on CalHire
There is no résumé screen, because the résumé is not the record.
- You declare your skills and prove them once. One AI-generated, cheat-resistant assessment — a skills test plus a text interview — produces a verified composite score from test, interview and role-fit, valid 90 days and portable across every application on the platform.
- If you upload a résumé, it only pre-fills your declared skills. It is never stored as your record and never shown as one. There is no résumé-match component in the score, at all.
- Employers see your verified score and skills — never your name, photo, age, school or employers. You are ranked on ability, under an opaque handle that is not derived from your personal information.
- Nothing is auto-rejected. Below-threshold applicants are flagged for human review, and a person makes every decision.
- Every application gets an answer. You receive a growth-framed report card showing where you stood and what to improve — enforced in the product, not optional.
- It is free for candidates, always.
See for candidates for the full journey, or /features for how the anonymity boundary works.
So: format your document properly, answer the screening questions carefully, and use the employer's vocabulary. Then be sceptical of anyone selling you protection from a robot whose existence they cannot document.
Frequently asked questions
- Do applicant tracking systems automatically reject résumés?
- Most do not silently auto-reject based on résumé content. What they commonly do is knock out applications that failed a screening question the employer configured — work authorisation, location, a required certification — and rank or filter search results so recruiters read some applications and not others. The effect can feel identical to rejection, but the mechanism is different and that matters for what you do about it.
- Is the "75% of résumés never seen by a human" statistic true?
- There is no traceable primary source for it. It circulates widely, is attributed inconsistently, and cannot be tied back to published data. What is true and less dramatic: many applications are never read, mostly because roles receive more applications than anyone reviews rather than because software silently discarded them.
- What résumé format works best with an ATS?
- A single-column document with standard section headings, real text rather than images, no tables or text boxes for important content, a common font, and a .docx or text-based .pdf. Complex two-column layouts and graphics genuinely lose information in parsing, and infographic CVs can parse to almost nothing.
- Should I add keywords from the job description?
- Use the employer’s actual vocabulary for things you genuinely have — if they say "Kubernetes" and you wrote "K8s", match their term. Do not stuff keywords, and never list skills you do not have. Recruiters read the document afterwards, and a mismatch between the keyword list and the described experience is very visible.