Do certifications actually get you past the resume screen?

Where the 75 percent figure comes from

On 28 February 2019 CNBC published a piece headlined "75% of resumes are never read by a human," which opened: "Three-fourths of all resumes never even get seen by human eyes, according to a study from job search services firm Preptel." That sentence is the load-bearing citation for a claim now repeated on thousands of career pages. Preptel sold resume optimisation. As of 28 August 2026 the domain preptel.com returns no DNS record, and no study, dataset, sample size or methodology from Preptel appears anywhere findable. A number that famous with no retrievable source is folklore, and the useful thing is to say so rather than repeat it. Two related figures circulate with the same problem. CNBC also wrote that "more than 95 percent of Fortune 500 companies use an ATS," attributed to "data collected by Jobscan," a company that sells resume scanning to job seekers. That figure comes from a company that sells resume scanning to job seekers, and it is not a survey you can inspect: CNBC links no dataset, sample or method behind it. The point is not that screening software is rare. The point is that the specific claim driving your anxiety, that three out of four applications are machine-deleted unread, has never been evidenced by anyone.

What actually happens when you press submit

Greenhouse publishes a description of its own pipeline, and it is the clearest account available from a vendor with something to lose by lying. Greenhouse says over 7,500 employers use its software and that its customers processed more than 150 million applications from over 40 million people in a single year. Its account of what happens next: "Once you click 'submit,' your application will go straight into a queue on a recruiter's dashboard in Greenhouse." Ninety-five percent of jobs are configured to send an automated acknowledgement email. Then, in Greenhouse's own words, "by and large, applications in Greenhouse are reviewed by real people, one by one, in the order received." On artificial intelligence specifically, Greenhouse states: "AI doesn't score or rank applications, nor does it make any decisions about whether or not you move forward in the process." That is one vendor, and other systems behave differently; Greenhouse itself acknowledges that "some companies use AI to analyze, score and auto-reject applications without any input from humans." But the default architecture is a filing cabinet with a search box attached, and the thing standing between you and an interview is the person holding the queue.

The only automatic rejection that exists in most systems

Greenhouse's support documentation for its auto-reject feature is worth reading closely, because it defines the boundary precisely. Auto-reject lets an employer "set up application rules so that, based on an applicant's answer to a question, they will automatically be rejected." The documentation then states the constraint: "Auto-reject application rules can only be configured for questions of the following type: Yes / No, Single-select, Multi-select." Your resume text is not one of those types. No rule can be built against the prose of your CV. The rejection fires on the form question, and the example Greenhouse gives is a driving job asking "Do you have a Class A Commercial Driver's License?" These are usually called knockout questions: work authorisation, willingness to relocate, minimum salary, a named licence. If you were rejected within minutes of applying, that is almost certainly what happened, and the trigger was an answer you typed into a dropdown, not a keyword you omitted. Greenhouse says over 50 percent of roles are configured to email you when an auto-reject rule fires. Silence, by contrast, usually means the opposite of rejection by machine. It means a recruiter filtered the queue to a shortlist and never reached your row.

What the search box does, which is where keywords half matter

The Harvard Business School and Accenture report Hidden Workers: Untapped Talent, published September 2021 by a team led by Joseph Fuller, surveyed 2,275 executives across the United States, the United Kingdom and Germany, alongside 8,720 workers. It found 63 percent of employers use a recruiting management system, rising to 69 percent among employers with more than 1,000 staff, and 75 percent in the United States against 58 percent in the UK and 54 percent in Germany. More than 90 percent of those employers use the system to filter or rank candidates: 94 percent for middle-skills roles and 92 percent for high-skills roles. The report then describes the mechanic exactly: "These systems allow employers to indicate specific requirements, such as possessing a specific academic or professional credentials (e.g., a bachelor's degree, a professional license, or certification) to filter the applicant pool and reduce it. The remaining applicants are then ranked based on other attributes or preferences." So a filter does exist. It runs on credentials and structured criteria first and ranks on keywords second, and a recruiter chooses what goes into it. Which means the useful question is not whether your resume is machine-readable. It is whether you hold the specific thing the recruiter is typing into the filter.

The filter that rejects most people is a tired human

Ladders published an eye-tracking study on 6 November 2018, updating one it first ran in 2012, which measured how long professional recruiters spend on a resume in the first pass. The finding: "The average initial screening time for a candidate's resume clocks in at just 7.4 seconds," up from the six-second average recorded in 2012. Ladders is a job board reporting on its own research, so treat it as one measurement rather than a settled fact, but it is the named, published source behind a number usually quoted with no attribution at all. Seven seconds is a scan for job titles, employer names, dates and layout. It is not a reading. ISC2's 2025 hiring trends survey of 929 cybersecurity hiring managers put a number on who does that scanning for entry- and junior-level positions: 53 percent said screening was shared between the hiring manager and HR, 35 percent said it was handled exclusively by IT or cybersecurity hiring managers, and 13 percent said it was handled solely by HR and software tools. In that sector, at that level, roughly seven applications in eight reach a human evaluator. The bottleneck is attention, not parsing.

The exception: when the posting names the credential

There is one situation where a certification changes what happens at the screen, and it is worth being precise about it. When a job description names a specific credential, that string becomes a filter term, and holding it moves you from the unranked pile into the searched set. Greenhouse describes the mechanism running in both directions: recruiters can set rules so that "you might be automatically advanced to a later stage if you meet certain criteria (if you hold a specific certification or are willing to travel, for example)." Employers often have a commercial reason for the requirement rather than a preference. AWS publishes the headcount requirements for its Services Partner tiers: Select tier requires 2 AWS Foundational Certified Individuals and 2 AWS Technical Certified Individuals, Advanced requires 4 and 6, Premier requires 10 foundational and 25 technical, of which at least 10 must be Professional or Specialty level. A consultancy at Advanced tier that loses certified staff loses the tier, the funding and the listing. When that firm posts a role naming AWS Certified Cloud Practitioner, it is filling a slot on a compliance sheet, and the credential is the entire qualification for that line. Same logic drives partner requirements across vendor ecosystems. If you are applying into those firms, the certificate is not decoration.

When the credential is a legal gate rather than a filter

A different case gets confused with resume screening and should not be. Some occupations cannot legally be practised without a licence, and no amount of resume optimisation substitutes. The Bureau of Labor Statistics states plainly for registered nurses: "Registered nurses must be licensed." That occupation held 3,465,400 jobs in 2025 with a May 2025 median wage of $97,550. An application from an unlicensed candidate is rejected because employing that person would be unlawful, and the knockout question on the form asking for a licence number is doing exactly the job it was built for. The same structure applies to any regulated practice with a state or national board behind it. If the credential you are missing is a licence, then the screen is not your problem and keyword advice is a distraction; passing the licensure exam is the whole task. If the credential you are missing is a vendor certification for an unregulated job, it is a filter term, valuable only when postings actually name it. Distinguishing between those two cases before you spend money is the single most useful thing you can do.

The test that answers this for your own search

Do not take a general answer when a specific one costs twenty minutes. Open the last twenty jobs you applied to. For each, record three things: does the posting name a specific certification by name; does the application form contain a yes-or-no or dropdown question about a licence, credential, authorisation or clearance; and did you receive a rejection email, an acknowledgement only, or nothing. Now count. If fewer than three of twenty name a credential, buying one will not change your response rate, and the problem lives in your titles, your target roles or the volume of applicants per posting. If eight or more name the same credential, you have found a filter term and a specific exam to sit, and you can price it against the salary of the roles naming it. If most of your rejections arrived within hours, look at the form answers rather than the resume, because that timing pattern matches Greenhouse's description of application rules firing on dropdown responses. Harvard Business School's report also found 88 percent of employers agree that qualified high-skills candidates get vetted out because they do not match the exact criteria in the job description, rising to 94 percent for middle-skills roles. Employers know the filter is crude. That is a reason to target postings whose criteria you meet, rather than to rewrite the same resume again.

What this does not fix

This post does not claim that applicant tracking software never rejects anyone automatically, and it does not claim that formatting is irrelevant. It claims only that the 75 percent figure has no published study behind it and that most automatic rejections fire on form answers rather than resume text. It cannot tell you which system a given employer runs, and some employers do use AI scoring that behaves differently from anything described here.

Where an exam fits

• AWS Cloud Practitioner, if the consultancies you are applying to name it, because their AWS partner tier depends on the headcount
• NCLEX-RN, if the credential blocking you is a licence rather than a filter term
• Salesforce Administrator, if the postings in your target list name it as a required certification
• CompTIA A+, if you are aiming at help desk roles where employers name an IT certification in the posting
• CompTIA Security+, if you are applying to federal contract work where the credential is a contractual condition of the seat

Common questions

Sources

1. 01Source given by CNBC for the claim that three-quarters of resumes are never seen by a human, and for its Fortune 500 ATS figure CNBC's 28 February 2019 piece, headlined "75% of resumes are never read by a human," states: "Three-fourths of all resumes never even get seen by human eyes, according to a study from job search services firm Preptel." The same piece states "More than 95 percent of Fortune 500 companies use an ATS ... according to data collected by Jobscan." Neither the Preptel study nor the Jobscan dataset is published or linked.
   CNBC Make It checked 2026-08-28
2. 02Employers using a recruiting management system to filter or rank candidates 63 percent of the 2,275 executives surveyed use an RMS, rising to 69 percent among employers with more than 1,000 workers (75 percent in the US, 58 percent UK, 54 percent Germany); more than 90 percent use it to initially filter or rank candidates, at 94 percent for middle-skills and 92 percent for high-skills roles. A parallel survey covered 8,720 workers.
   Harvard Business School and Accenture, Hidden Workers: Untapped Talent (Fuller et al., September 2021) checked 2026-08-28
3. 03Employers agreeing that qualified candidates are vetted out for not matching exact job description criteria 88 percent for high-skills candidates, 94 percent for middle-skills, from a survey of 2,275 executives across the US, UK and Germany
   Harvard Business School and Accenture, Hidden Workers: Untapped Talent checked 2026-08-28
4. 04Question types on which an applicant tracking system can be configured to auto-reject Yes/No, single-select and multi-select application questions only; resume text is not one of them
   Greenhouse Support documentation, Auto-reject checked 2026-08-28
5. 05Greenhouse's own account of how applications are handled and whether AI ranks them Over 7,500 employers, 150 million applications from more than 40 million people in a year; 95 percent of jobs auto-send an acknowledgement; applications "reviewed by real people, one by one, in the order received"; "AI doesn't score or rank applications"
   Greenhouse checked 2026-08-28
6. 06Roles configured to email the applicant when an auto-reject rule fires Over 50 percent
   Greenhouse checked 2026-08-28
7. 07Average time recruiters spend on the initial screen of a resume 7.4 seconds, up from six seconds measured by the same study in 2012
   Ladders, Inc. press release, 6 November 2018 checked 2026-08-28
8. 08Who screens applications for entry- and junior-level cybersecurity roles 53 percent shared between hiring manager and HR, 35 percent exclusively by IT/cybersecurity hiring managers, 13 percent solely by HR and software tools, from 929 hiring managers surveyed
   ISC2 2025 Cybersecurity Hiring Trends Report checked 2026-08-28
9. 09Certified individuals required to hold each AWS Services Partner tier Select tier: 2 AWS Foundational and 2 AWS Technical Certified Individuals; Advanced: 4 foundational and 6 technical, minimum 3 at Professional or Specialty level; Premier: 10 foundational and 25 technical, minimum 10 at Professional or Specialty level
   Amazon Web Services checked 2026-08-28
10. 10Licensure requirement and size of the registered nurse occupation "Registered nurses must be licensed"; 3,465,400 jobs in 2025 and a May 2025 median wage of $97,550
    U.S. Bureau of Labor Statistics, Occupational Outlook Handbook checked 2026-08-28
11. 11Current DNS status of preptel.com, the firm CNBC credited with the 75 percent study Google Public DNS returns Status 3 (NXDOMAIN) for preptel.com, meaning the domain does not resolve; no Preptel study, sample size or method is retrievable anywhere
    Google Public DNS resolver query checked 2026-08-28

Related