RESEARCH & SOURCES

What the study reports: In this group of 25, 92% of systems (23/25) did not auto-reject for formatting, content, or design. Only 2 of 25 were configured to auto-reject on content-based rules. 84% said their system relies on knockout questions. 52% said applying early improves a candidate’s chances; 36% said it makes little difference. 8 of 25 (32%) recommended proactive LinkedIn outreach or referrals as a way to stand out.

Caveats the book attaches: Enhancv sells resume software — the study appears on its blog alongside its products. The sample is small (25 interviews) and U.S.-only. The study itself calls it small and says it shows consistent themes, not nationally representative statistics. Treat it as moderate-weight evidence for themes; low weight for prevalence figures.

How READ BY HUMAN uses this: Chapter 1 uses this study to distinguish what the evidence actually supports (knockout questions and volume are common explanations; content-based auto-rejection appears uncommon in this group) from the popular but unsourced claim that “bots reject 75% of resumes.” Chapters 4, 6, and 7 use the recruiter descriptions of what they valued in resumes and the referral recommendation.

What the data shows: 1.9 million people were long-term unemployed (27 weeks or more) in August 2026, representing 27.0% of all unemployed. The March 2026 figure was 1.8 million (25.4%), up from 1.5 million (21.3%) a year earlier.

How READ BY HUMAN uses this: Cited as U.S. government labor market data to establish the duration problem mid-career job searches face. The book uses the BLS figures only as reported — it does not extrapolate beyond the headline numbers.

What the documentation establishes: These are FACTS that the features exist as described — not interpretations. Greenhouse’s auto-reject ties to subscription tier; Oracle Recruiting uses response scores with negative scores for disqualifying answers; Taleo Enterprise lets each answer be set to pass, disqualify, or verify. Auto-advance to later interview stages is also a documented feature.

Caveats the book attaches: Vendor documentation establishes that a feature exists and how it is described — it says nothing about how often employers choose to configure or use it. These are facts about what the tools can do, not about what recruiters typically do.

How READ BY HUMAN uses this: Chapter 1 cites these to establish that content-based auto-rejection, knockout questions, and structured search are real features employers can configure — not hypothetical. The Worksheet 1 (ATS Reality Check) asks candidates to treat every stated hard requirement as a potential knockout.

What the documentation establishes: LinkedIn Recruiter supports keyword and Boolean search, advanced filters (job titles, location, companies, skills, schools, industries), must-have / can-have options, and an “Open to work” filter. These are vendor-documented features.

How READ BY HUMAN uses this: Chapter 4 uses this to explain why using recognizable titles, skills, and tools in plain language makes a candidate findable in recruiter search — not just readable by humans. The book notes that visibility settings for the “Open to work” signal were not verified against LinkedIn’s member help pages and directs readers to check current settings themselves.

What the survey reports: 58% of hiring managers said they received AI-generated resumes or cover letters; 34% found AI-generated LinkedIn or social profiles; 82% expressed concern about candidates’ use of AI; 86% agreed AI will create authenticity-verification challenges.

Caveats the book attaches: Resume Genius sells resume-building tools. The sample size is internally inconsistent (1,500 vs. 1,000 stated on the same page). Treat as moderate-weight evidence for direction of concern; not a precise prevalence measure.

How READ BY HUMAN uses this: Cited in the AI philosophy sections to support the direction of concern — that hiring managers are becoming more skeptical of AI-generated content and more focused on authenticity — not as a precise frequency measure.

What the study finds: A field experiment on an online labor market with nearly half a million jobseekers. Treated jobseekers were hired 8% more often. No evidence employers were less satisfied. The writing service did not provide whole paragraphs and could not be prompted.

Caveats the book attaches: The study was conducted on an online labor market, not in corporate hiring. The publisher’s page notes funding from the marketplace. The writing tool was constrained — it was not a large language model you could prompt freely.

How READ BY HUMAN uses this: This study is the closest thing to controlled evidence that AI writing assistance can improve outcomes without deceiving employers. The book cites it to show AI assistance is not inherently dishonest — the constraint is that AI must add no facts, not that it can add no words. The Prompt Library’s design (every prompt restricts the AI to facts you supply and ends with a verification step) is built directly on this distinction.

What the study finds: Structured interviews emerged as the top-ranked selection procedure in this revised analysis. Mean validity estimates for many selection procedures were reduced by roughly 0.10–0.20 relative to earlier summaries (including Schmidt & Hunter’s widely cited earlier work). Selection procedures remain useful, but predictor–criterion relationships are considerably lower than previously thought.

Caveats the book attaches: The book cites the ranking and the direction of the revision only — not specific coefficients — because the abstract text did not state them and secondary sources that did were not authoritative enough to cite. Sackett et al. (2022) specifically revises the Schmidt & Hunter estimates downward; this is why the book cites the more recent paper.

How READ BY HUMAN uses this: Chapter 8 uses the structured-interview ranking to explain why the Interview Story Bank (Worksheet 11) and Prompt P16 (Evidence-to-Requirement Mapper) prepare candidates to speak to stated criteria — because hiring managers using structured rubrics are the ones making the decisions that matter most.