Is a robot really throwing your CV away?

Myth 1: "75 per cent of CVs are never seen by a human"

Verdict: false. There is no study behind the number.

This is the most repeated figure in the whole subject, and it has been traced back to its source more than once. In October 2020 Christine Assaf, an HR professional, heard it at a presentation and followed the citations: the speaker cited the CV-writing company TopResume, TopResume cited a Forbes article, and the Forbes article cited nobody. The outlets that did name a source, CIO and CNBC, attributed it to "job search services firm Preptel", a CV-writing company that closed in August 2013 and never published a study, a dataset or a method (Ask a Manager, reprinting Christine Assaf (opens in a new tab)).

In July 2026 two Canadian academics, Leda Salazar de Leon of Mount Royal University and Mehnaz Rafi of the University of Calgary, reached the same conclusion in The Conversation: the figure "originated from a 2012 sales pitch by Preptel, a resume optimization company that went out of business the following year. No methodology was ever published" (The Conversation (opens in a new tab)).

The article that started the wave, Lauren Weber's "Your Résumé vs. Oblivion" in the Wall Street Journal in January 2012, did not say it either. Asked on NPR to explain the systems, Weber described software that will "score and rank the candidates" so that a recruiter can open the job and look at the top 25 or 50. Her worst case was not the bin. It was "the bottom of the ranking system" (NPR, Talk of the Nation, February 2012 (opens in a new tab)). The number drifts between 70, 75 and 88 per cent depending on who is retelling it, which is what happens when there was never an original to check.

Myth 2: "The ATS rejects your CV automatically"

Verdict: mostly false, and the part that is true is worse than the myth.

In the official documentation of Greenhouse, Workday and iCIMS, the only automatic rejection is triggered by your answers to the employer's own questions on the application form: do you hold the licence, are you allowed to work here, can you work weekends (Greenhouse, Auto-reject (opens in a new tab); Workday Administrator Guide (opens in a new tab); iCIMS (opens in a new tab)). None of them describes a hidden score that rejects a CV. Where AI matching exists, Greenhouse states that it sorts applicants into groups and "does not automatically disposition any candidates nor does it make any hiring decisions" (Greenhouse, Talent Matching FAQ (opens in a new tab)).

Now the true part. The filters that recruiters set up are real, and they are blunt. In Harvard Business School's 2021 survey of 2,275 executives in the United States, the United Kingdom and Germany, almost half of the companies automatically screened out any candidate with a gap in employment of more than six months, and 88 per cent of the executives said qualified high-skills candidates were vetted out because they did not match the exact criteria (Hidden Workers: Untapped Talent (opens in a new tab)). A person chose that rule, and a person pressed reject on the result. Assaf put it in one sentence: "A human rejects the applicants; a computer ranks applications based upon data provided by a human."

Myth 3: "An AI reads your CV"

Verdict: mostly false, for now.

In January 2024 the Society for Human Resource Management surveyed 2,366 HR professionals in the United States. Only 26 per cent of their organisations used AI for any HR activity at all; of those, 64 per cent used it somewhere in recruiting, interviewing or hiring, and about one in three of that group used it to review or screen CVs (SHRM, 2024 Talent Trends (opens in a new tab)). Multiply it through and AI was reading CVs at roughly one organisation in twenty. Everywhere else, the software was running rules and searches that a person had set up. The share is growing, and the European Union's AI Act already classes recruitment AI as high-risk, but if you are picturing a machine that understands your CV, you are picturing the exception. I have described what the typical system does, step by step, in the first article in this series (opens in a new tab).

Myth 4: "An ATS cannot read PDFs, and a two-column layout gets you rejected"

Verdict: half true, and the true half is about layout, not file type.

The file type first. Greenhouse accepts .doc, .docx, .pdf, .rtf and .txt (Greenhouse, Supported formats (opens in a new tab)). iCIMS extracts text from PDF, Word, OpenDocument, RTF, HTML and even image files (iCIMS Developer Community (opens in a new tab)). Textkernel, whose parser sits inside many of these systems, reads more than 70 formats, which it says "cover nearly 100% of the documents that we process in practice" (Textkernel, Supported File Formats (opens in a new tab)). A PDF exported from Word or Google Docs is text, and a parser reads it. What a parser cannot read is a picture of text: a scanned CV, or a design tool's export where the letters are pixels. That needs optical character recognition, which Textkernel sells as an add-on and warns takes "considerably longer to parse".

Now the layout. Textkernel's own documentation warns that columnar layouts confuse the parser and that some sections are "impossible to parse at a granular level" (Textkernel, Parser Output (opens in a new tab)). The result is not a rejection. It is a record in which your last job is attached to the wrong employer, or your skills sit in no field at all, and that record does not come up when the recruiter searches. Nothing tells you. My icons and my two columns were doing exactly this, for a year.

Myth 5: "Hide the keywords in white text and the software will rank you higher"

Verdict: false, and it costs you the job.

A parser does not see colour. It extracts the text and stores it as plain text, and that plain text is what the recruiter opens. Jobscan, a company that sells CV scanning, shows what that looks like: a parsed CV with "Accounting Accounting Accounting" running across it, which is "easy to spot once the resume text has been parsed into the system" (Jobscan, November 2024 (opens in a new tab)). The recruiter sees the trick before they see your name.

In 2023 the trick came back in a new form, aimed at AI screening: paste the whole job description into the CV in white font so that the model matches you to it (Bruce Schneier on a Washington Post report, August 2023 (opens in a new tab)). Same trick, same plain-text view, same recruiter. And in 2023, as Myth 3 shows, the "AI" reading most CVs was still a search box. I built Adapt to Success on the opposite principle: it changes structure and wording, never facts, and never anything a reader cannot see.

Myth 6: "You need a 75 per cent keyword match"

Verdict: the number does not exist. The words do.

The match percentage comes from CV-scanner tools, not from the employer's software. Jobscan, which recommends aiming for 75, says so on its own scanner page: "when you submit your resume to a job, the ATS doesn't score your resume. The match rate in our scanner is a visualization tool" (Jobscan, Resume Scanner (opens in a new tab)). A recruiter never sees that number, and nothing in the systems I have read about computes it.

What actually happens is plainer. The recruiter types words from the job advert into a search box, and the system can rank the records by how closely they match the requirements (Indeed Career Guide (opens in a new tab)). Two things follow. First, the software has become more forgiving about synonyms: Textkernel's parser normalises more than 250,000 skill synonyms into about 13,000 skills across more than 20 languages (Textkernel, Tx Platform FAQ (opens in a new tab)), so the "numerical modeler" who was invisible to a search for "statistician" in Weber's 2012 example has a better chance today. Second, a search box still finds what is typed, and the Harvard survey's 88 per cent of executives who see qualified people vetted out on exact criteria are describing precisely that. Use the words of the advert where they are true of you. Ignore the percentage.

Myth 7: "Every company uses an ATS"

Verdict: half true.

Among the largest employers it is nearly true. Jobscan looked at the job pages of all 500 companies on the 2026 Fortune 500 list and detected an ATS at 487 of them, 97.4 per cent (Jobscan, August 2026 (opens in a new tab)). That is a vendor's count, made by reading career pages, but it is a count, not a guess. Across employers in general the picture is different: Harvard Business School's survey found 63 per cent using a recruiting management system across the United States, the United Kingdom and Germany, with 75, 58 and 54 per cent by country, and 69 per cent among employers with more than 1,000 staff (Hidden Workers: Untapped Talent (opens in a new tab)). The Conversation's authors add that small businesses are especially unlikely to rely on one (The Conversation (opens in a new tab)). So: a large employer, almost certainly. The company with twelve staff on the high street, probably not. The next article in this series looks at the figures region by region.

Myth 8: "An ATS-friendly template is all you need"

Verdict: false.

A template makes you readable. It does not make you relevant. After the parser comes a person, and the person is quick: in 2018 the careers site Ladders measured recruiters' first look at a CV with eye-tracking equipment and got an average of 7.4 seconds (HR Dive on the Ladders study (opens in a new tab)). The Harvard report describes what happens before that look: the software "largely relies on 'negative' logic to winnow the applicant pool", removing people for what they lack rather than considering what they have. A perfectly parsed CV that says the wrong things is simply found, and then declined, with the small consolation that this time a person saw it. Structure gets you into the search results. Content gets you the call. There are four before-and-after examples of both on the ATS Explained page (opens in a new tab).

What is actually true

• The software stores, searches and ranks. People reject.
• Automatic rejection comes from your answers to the employer's questions, and from filters a recruiter set, not from a hidden CV score.
• Your CV must be readable by a parser: one column, real text rather than images, standard section headings, and dates attached to the jobs they belong to.
• Your CV must contain the words the advert uses, where they are true of you.
• Then a person reads it for about seven seconds, so it must be clear to a person too.

What I wish I had known

I spent a year adding icons to a document that software was reading as text, and polishing columns that a parser was reading as noise. Nobody was throwing my CV away. It was sitting in a database, looking like the CV of someone who had never done what the advert asked for. The first two points on the list above are the ones that software can fix, and they are what Adapt to Success does: it keeps your facts and rebuilds the structure and the wording so that a parser and a search box can read them. The last three are yours.