No published study shows that applicant tracking systems (ATS) reject 75% of resumes before a person sees them. The figure goes back to 2012 marketing from Preptel, a company that sold resume optimisation, and no methodology was ever published. What is well documented is narrower: employers use software to filter and rank candidates, and most employers in one large survey believe it screens out qualified people.
Who this is for: job seekers who’ve been told to “beat the ATS”, and anyone who has repeated the 75% figure and wants to know where it came from.
Where does the “75% of resumes are rejected by ATS” statistic come from?
The earliest version we could find is Preptel’s launch press release, distributed by Marketwire on 31 January 2012. It says that “over 75 percent of resumes are discarded for using the wrong word or formatting issues” (press release copy). In the same release, the company describes its own product as a resume optimisation service.
Look at what the release doesn’t contain: no survey, no sample size, no definition of “discarded” and no method. It’s a claim from a company selling the fix for the problem it describes. Researchers who have traced the citation chain report the same dead end, and they note that Preptel closed in 2013 without publishing any data behind the number (one write-up of the trace).
From there, the figure went around the internet. Articles quoted other articles, the number drifted between 70%, 75% and 88%, and the original context dropped away. That doesn’t prove the real rate is low. It means nobody knows the real rate from this source, because there was never a measurement.
What does the Harvard Business School “Hidden Workers” report actually say?
The figure you’ll often see next to the 75% is 88%, and it does have a source: Hidden Workers: Untapped Talent by Joseph B. Fuller, Manjari Raman, Eva Sage-Gavin and Kristen Hines, published in 2021 by Harvard Business School’s Project on Managing the Future of Work together with Accenture (report page, PDF).
What the report found, in its own terms:
- It is based on surveys of more than 2,250 employers and more than 8,000 workers in the US, the UK and Germany (Accenture summary).
- More than 90% of employers said they use recruiting software to initially filter or rank candidates: 94% for middle-skills roles and 92% for high-skills roles.
- 88% of employers agreed that qualified high-skills candidates are screened out because they don’t match the exact criteria in the job description. For middle-skills roles the figure was 94%.
| Claim | Source | Method published? | What it measures |
|---|---|---|---|
| “75% of resumes are rejected by ATS” | Preptel press release, 2012 | No | Nothing measured |
| “88% of employers say qualified candidates get screened out” | Hidden Workers, HBS & Accenture, 2021 | Yes | What employers believe |
| “Over 90% of employers use software to filter or rank” | Hidden Workers, HBS & Accenture, 2021 | Yes | Reported use of recruiting software |
This is useful evidence, but read it carefully. It’s employers saying what they believe their process does. It’s not a measured share of resumes rejected by software. It tells you the problem is real, but not how big it is for your application.
The report also explains how qualified people get screened out: rigid proxies such as requiring a degree, or excluding anyone with a gap in full-time employment, whatever their other qualifications. These are rules people configure, which is the important part.
What does an ATS actually do?
An applicant tracking system is mainly a database and a workflow tool. In most setups it:
- Stores your application and parses your resume into fields (name, roles, dates, skills).
- Lets recruiters search and filter that database, for example “Python” and “Berlin”.
- Applies employer-set rules, such as knockout questions (“Are you authorised to work in the UK?”) or a minimum years-of-experience question.
- Sometimes ranks or scores candidates, depending on the product and how it’s set up.
- Tracks the process: stages, interview scheduling and rejection emails.
So when software rejects you automatically, the cause is much more often a knockout answer than a bad keyword density. When nobody replies, the usual reason is volume: a recruiter with hundreds of applicants who never got to yours, or read it and moved on.
So what actually gets resumes screened out?
In our experience from the hiring side, these are the common reasons, roughly in order:
- Knockout criteria: location, work authorisation, required licence or clearance, minimum experience.
- Level mismatch: a mid-level resume for a staff role, or the other way round. Our guide to engineering levels on a resume shows what each level is expected to look like.
- The match isn’t visible: the relevant experience is there but buried on page two.
- Parsing problems: text inside images, complicated multi-column layouts, or headers the parser can’t recognise. These are worth fixing, but they’re the easy part.
- A human decision: someone read it and chose other candidates.
None of these is fixed by stuffing keywords.
How should you write a resume that both software and people can read?
Keep the format boring
Use one column, standard headings (“Experience”, “Education”, “Skills”), real text instead of images, and consistent dates. Export a PDF that has selectable text. That covers most parsing problems.
Answer screening questions honestly
Knockout questions are where automatic rejection really happens. If you don’t meet a hard requirement such as work authorisation, a keyword won’t change that. If you meet it, make sure your answer is clear.
Make the match obvious to a person
Put the 3–5 experiences most relevant to the role first. Describe the scope in words the reader recognises. This is the core of tailoring a resume to a job description without lying.
Use the posting’s words only where they’re true
If you did “release management” and called it “shipping”, use their term. Don’t add tools you haven’t used. Someone will ask about them in the interview.
Don’t invent numbers to “score higher”
Adding made-up percentages doesn’t get you past the software and does hurt you with people. When you don’t have a number, describe the scope and the before and after. See resume bullets without metrics.
Software-and-human readability check
0 of 8 done
It depends on the company
- Big tech and large employers get huge volumes, use ranking and filters heavily, and set knockout rules strictly. A clean format and clear level signals matter most here.
- Scale-ups often use lighter ATS setups, and recruiters skim fast. Put the match on page one.
- Early startups may have founders reading every application themselves. Being specific beats polish.
- Enterprises and the public sector often have formal criteria lists. Address each listed requirement explicitly, and honestly.
What not to do
- Don’t hide keywords in white text. Many systems show the parsed text to recruiters, and it looks bad when they see it.
- Don’t pay for “ATS score” tools that promise a pass rate. Nobody outside the employer knows that employer’s settings.
- Don’t repeat the 75% figure as fact, including in your own posts. Say what it is: an unsourced vendor claim from 2012.
FAQ
Is the 75% ATS rejection statistic true?
There's no published evidence for it. It goes back to a 2012 claim by Preptel, which sold resume optimisation, and no method or data was published.
What is the 88% statistic about ATS?
It comes from the 2021 Harvard Business School and Accenture report Hidden Workers: Untapped Talent. 88% of surveyed employers agreed that qualified high-skills candidates are screened out for not matching the exact job criteria. It measures what employers believe, not a rejection rate.
Do ATS automatically reject resumes?
Sometimes, mostly through knockout questions and rules the employer configures, such as location or work authorisation. Ranking and filtering also affect who gets read first.
Should I optimise my resume for ATS keywords?
Use the employer's terms for things you really did, and keep the format simple. Beyond that, write for the person who will read it.
Does a PDF or Word file work better with ATS?
Most modern systems handle a text-based PDF well. If a posting asks for a specific format, use that.