What Gets a Resume an Interview: Evidence on ATS, Length and Bias

A resume is only the first screen, and the evidence on what passes it is thinner than the advice industry implies. Recruiter interviews suggest bots rarely reject resumes outright, while audit experiments show that employers also react to signals unrelated to qualifications.

A resume does not win an interview on its own. It is the first screen in a longer process, and the best available evidence suggests that a person, not an algorithm, usually makes the call on it. That person works fast, reads for fit, and, according to decades of field experiments, can also be swayed by signals that have nothing to do with qualifications.

Much of the popular advice on resumes, from the claim that bots reject 75% of applications to the rule that one page is always enough, rests on studies run by companies that sell resume tools or writing services. Separating what is documented from what is repeated matters because job seekers spend real time on formatting and keyword tricks that may do little.

Who actually screens a resume

An applicant tracking system, or ATS, is software that stores, sorts, searches and manages applications. A widely repeated belief holds that such systems automatically discard most resumes before a human sees them. The evidence gathered so far does not support that.

What recruiters say about automatic rejection

The most direct data come from Enhancv, a resume-tool company that interviewed 25 US recruiters. Twenty-three of them (92%) said their systems do not automatically reject resumes because of formatting, content or design. Only 2 of 25 (8%) said their ATS auto-rejects for reasons other than knockout questions. According to HR.com's republication of the release, the fieldwork ran from September to October 2025 and covered companies with 120 to more than 50,000 employees.

The same study found that 56% of recruiters disregard automated fit percentages in favor of manual review, and 52% said that applying within 48 to 72 hours significantly raises the chance of being reviewed before shortlists are finalized. Those are self-reported views from a small, non-random sample. Enhancv also traces the 75% rejection claim to an unsourced 2012 sales pitch, though that origin was not independently verified.

A second vendor, Huntr, says its analysis of more than 1.7 million applications plus recruiter interviews found that no major ATS auto-rejects resumes and that no universal ATS score exists. The methodology was not visible in the material reviewed. A Substack post, relaying an analysis by the vendor CV Compiler, puts the share of tech resumes that fail at the parsing stage at about 3%, but that is a secondhand vendor figure and a weak basis for firm conclusions.

The real exception: knockout questions

Automatic rejection does exist in one form. A knockout question is an employer-written eligibility question with an automatic consequence. According to a ResumeAdapter vendor blog, employers can set rules tied to work authorization, sponsorship, location, a required license or minimum years of experience. That is consistent with Enhancv's finding that all of its recruiters used knockouts where they were present.

The practical reading is that a resume is unlikely to be thrown out for a two-column layout, but an answer that trips an employer's eligibility rule can end an application without anyone reading the document.

A dissenting concern

The picture is not fully settled. A HiringThing blog cites a Harvard Business School and Accenture report saying that ATS algorithms could be responsible for excluding whole segments of viable candidates. The primary report was not located, so the claim could not be checked. Every source behind the low-rejection finding is a vendor with a commercial interest, and no independent audit of how employers configure their systems was found. The defensible summary is that auto-rejection appears to be rare outside knockout rules, not that it never happens.

The seconds-long scan and what it implies

If a person does the screening, how long does the person spend? The best-known answer is The Ladders' 2018 eye-tracking study, which reported an average initial screen of 7.4 seconds. An earlier version of the study found 6 seconds, during a recession when recruiters were overwhelmed with applicants.

The number deserves caution. The Ladders sells resume-writing services, and secondary sources say the full methodology was never published in a peer-reviewed journal. Those secondary accounts say 30 recruiters were tracked over 10 weeks, but one critic notes that the report does not state how many recruiters took part, so the sample size cannot be confirmed from the primary document. The figure also describes only the first fit or no-fit scan, not total reading time.

The direction of the finding is more trustworthy than the number. The Ladders reported that a recruiter who is engaged spends as much time on a second page as on the first, and that keywords work best when they appear in context. CareerOneStop, a Department of Labor-sponsored site, says a reviewer who has trouble scanning a resume quickly is likely to dismiss even a qualified candidate. It lists spelling errors, too many fonts, inconsistent punctuation and a messy layout among reviewers' top frustrations.

What experiments with fake resumes show

The strongest causal evidence comes from audit, or correspondence, studies. Researchers send matched fictitious resumes to real job ads, randomly varying one attribute, and count callbacks, meaning requests for an interview or more information. Because the variation is random, differences in callbacks can be attributed to the attribute. The limits are real, though: these studies capture only the first screening stage, depend on the occupations sampled, change only a few features, and say nothing about who is eventually hired.

Names, race and the limits of resume craft

In an NBER working paper, Marianne Bertrand and Sendhil Mullainathan sent fictitious resumes to newspaper ads in Chicago and Boston, using names that signal race. Applicants with White-sounding names needed about 10 resumes to get one callback, versus about 15 for applicants with African American-sounding names, a 50% gap. The paper also found that higher-quality resumes drew more callbacks. A hint in the excerpt reviewed suggests the payoff from quality was smaller for African American names, but that point could not be verified and is left as an open question here.

A 2017 PNAS meta-analysis by Lincoln Quillian and colleagues pooled US field experiments. Its core analysis, covering fieldwork since 1989 and 24 studies, found no change in discrimination against African Americans over that period. The excerpt reviewed was cut off on the results for Latinos, so those are not reported here. Together, the two studies make it highly likely that name-based discrimination appears in audit data. How large it is today is less certain.

Gender and the broader pattern

The gender evidence is mixed. A preregistered meta-analysis of 44 years of field audits asked whether gender discrimination depends on how typical a job is for one gender, and reported an odds ratio of 0.75 (95% confidence interval 0.68 to 0.83) for female-typed jobs. The excerpt did not make the direction of the comparison clear, so the number is best read as evidence that job type matters. A separate meta-analysis in Sociological Science notes that some scholars find no gender discrimination in US audit studies while others find that it persists.

A PLOS ONE meta-analysis of 21 published correspondence studies, covering 592 social signals, found that perceived warmth and competence predicted callback disparities for race and gender, which names signal indirectly. In plain terms, employers appear to respond to impressions as well as to listed qualifications.

Which credentials moved callbacks

The only experiment found that tests which credentials raise callbacks is an arXiv preprint, a large-scale résumé audit that has not been peer reviewed. It reported that two credentials substantially raised callback rates: an internship emphasizing sales and client interaction, and a combination of programming and data-analysis skills. Base callback rates for the White-male reference group varied widely by occupation, from 5% in business and financial operations to 10% in office and administrative support, 12% in management and 25% in sales.

The preprint also reported that, in the four occupation groups with significant gaps, callback rates for the disadvantaged groups were 18% to 28% below the White-male rate. Credentials that raise callbacks narrowed those gaps only where evaluation is structured. The authors, design and sample size were not visible in the material reviewed, so this finding carries low confidence until it is peer reviewed or replicated.

Length, format and tailoring

The one-page rule is the most argued-over piece of resume advice, and the evidence behind either side is modest. No peer-reviewed study or real-world callback experiment on resume length turned up.

The two-page simulation and its misreporting

ResumeGo, a paid resume-writing service, reported that hiring managers and recruiters were 2.3 times as likely to favor two-page resumes over one-page versions. The design was a simulation in which half the resumes ran one page and half ran two. The page's metadata dates it to January 2019. Secondary sources, which the primary page did not confirm, put the ratio at 1.4 for entry-level roles and 2.9 for managerial roles.

Several websites misdescribe the study as showing 2.3 times more callbacks, and some misdate it to 2024. It measured recruiter preference in a simulation, not real interview invitations, and it dates from 2018.

Surveys and official advice

Secondary reports of recruiter surveys point the same way, though the primary documents were not located. An Accountemps poll of 150 executives in 2018 found 52% calling one page ideal for a staff-level resume, down from 73% a decade earlier. A 2024 Criteria Hiring Report, relayed by Fortune, found that 53% of recruiters expect two-page resumes. Surveys capture stated preferences, not hiring outcomes.

Official guidance is more conservative. CareerOneStop's resume page says to include only the experiences that make a candidate a good fit and to limit the resume to one page. Another CareerOneStop page says federal resumes are typically longer than the 2-page private-sector resume, with 4 to 6 pages common. The two statements do not contradict each other because the audiences differ. A reasonable reading is that a one-page limit for early-career applicants rests on official advice and convention, not on experiments, and that two pages are not shown to be penalized for experienced candidates.

Tailoring to the posting

US federal sources consistently recommend matching the resume to the job. The Department of Labor advises tailoring a resume to the position's requirements and avoiding misspelled words and bad grammar. The NIH Office of Human Resources advises mirroring the announcement's wording; its example is that if a job mentions developing something, using the word develops can show an ability to work independently.

That guidance is aimed at federal applications, and no experiment testing how much tailoring raises callbacks was found. Tailoring is sound practice on the strength of consistent official advice; its size of effect is unresolved.

When software does the ranking

Some employers are experimenting with language models to rank applicants, and research suggests the tools can inherit bias. Kyra Wilson and Aylin Caliskan of the University of Washington audited three text-embedding models in a peer-reviewed paper at the AIES 2024 conference, and Tech Xplore's report says they found significant racial, gender and intersectional bias in how the models ranked resumes.

According to a University of Washington release, white-male-associated names were favored 85% of the time and female-associated names only 11% of the time, and white-male names were always favored over Black-male names. The test used 500 real job listings across nine occupations. Reports describe the resume sample inconsistently, and the original paper says 120 frequency-controlled names were used.

The study tested research retrieval models, not commercial ATS products, so it shows what such models can do, not what deployed hiring software does. That sits alongside the recruiter reports that automated scores are often disregarded. How much of real screening runs through such rankings remains unmeasured in the material reviewed.

What it means for the next application

Taken together, the sources support a modest, practical picture. The first reader is probably a person working quickly, so a clean layout, clear job titles and consistent formatting serve the reader. Matching the posting's terms is endorsed by federal guidance, and answering eligibility questions accurately matters because knockout rules are the one documented form of automatic rejection.

The audit studies add a harder truth. Resume quality helps, but a better document does not erase the effect of signals such as names, and one preprint suggests credentials narrow gaps only where evaluation is structured. Resume craft is a lever applicants control; it is not the only lever shaping who gets called.

What to watch next is independent evidence: peer review of the credentials preprint, audits of real screening software, and field experiments on length and tailoring. Until then, confident rules about resumes deserve the same scrutiny as the 75% rejection claim.

Frequently asked questions

Do applicant tracking systems automatically reject resumes?

Recruiter interviews suggest it is rare outside knockout questions. In Enhancv's 2025 study of 25 US recruiters, 23 said their systems do not auto-reject for formatting, content or design. The evidence is self-reported and vendor-run, and no independent audit exists, so rare does not mean never.

How long do recruiters look at a resume?

The Ladders' 2018 eye-tracking study reported an average initial screen of 7.4 seconds, up from 6 seconds in an earlier version. The company's methodology was not published in a peer-reviewed journal, and the figure covers only the first fit or no-fit scan, so it is best treated as directional.

Should a resume be one page or two?

No real-world experiment on length was found. CareerOneStop advises one page, while a ResumeGo simulation and recruiter surveys lean toward two pages for experienced workers. The ResumeGo study measured recruiter preference, not actual callbacks.

Does tailoring a resume to a job posting help?

The Department of Labor, NIH and CareerOneStop recommend matching a resume to the position's requirements and wording. No experiment tested how much tailoring raises callback rates, so the benefit is recommended practice, not a measured effect.

Can a strong resume overcome hiring bias?

Audit studies show higher-quality resumes draw more callbacks, but they also show gaps tied to names and other signals. A meta-analysis found no change in discrimination against African Americans since 1989, so resume improvements help without removing bias.

Sources

1. Does the ATS Reject Your Resume? 25 Recruiters Explain What Really Happens , Enhancv
2. ATS Rejection Myth Debunked: 92% of Recruiters Confirm Applicant Tracking Systems Do NOT Automatically Reject Resumes , HR.com
3. How Applicant Tracking Systems Actually Work in 2026 , Huntr
4. Do ATS Automatically Reject Resumes? What Really Triggers It , ResumeAdapter
5. Applicant Tracking Systems Aren't Excluding Job Applicants—People Are , HiringThing
6. Why do recruiters spend only 7.4 seconds on resumes? , The Ladders (2018-01-01)
7. Does your resume speak to employers , CareerOneStop
8. Are Emily and Greg More Employable than Lakisha and Jamal? (NBER Working Paper 9873) , National Bureau of Economic Research
9. Meta-analysis of field experiments shows no change in racial discrimination in hiring over time , PNAS (2017-01-01)
10. On the trajectory of discrimination: A meta-analysis and forecasting survey capturing 44 years of field experiments on gender and hiring decisions , ScienceDirect (2023-01-01)
11. A Meta-Analysis of U.S. Audit Studies of Gender Bias in Hiring , Sociological Science
12. Perceived warmth and competence predict callback rates in meta-analyzed North American labor market experiments , PLOS ONE (PMC) (2024-01-01)
13. Hiring Discrimination and the Task Content of Jobs: Evidence from a Large-Scale Résumé Audit , arXiv (2026-01-01)
14. Settling the Debate: One or Two Page Resumes , ResumeGo (2019-01-22)
15. Resumes , CareerOneStop
16. Federal employment , CareerOneStop
17. Tips for Writing a Federal Resume , US Department of Labor
18. Writing a Federal Resume , NIH Office of Human Resources
19. AI tools show biases in ranking job applicants' names according to perceived race and gender , Tech Xplore (2024-10-01)
20. UW research finds racial and gender bias in AI tools ranking job applicants' names , University of Washington (2024-12-12)
21. Applicant tracking system (ATS) rejected my resume: Is it true? , RelocateMe (Substack)

This article was drafted with the help of AI and reviewed by Jaroha editors before publication.