The shifted burden of search: how digital hiring platforms reallocate screening labor to job seekers

1 Introduction

The contemporary job search is conducted predominantly through digital platforms, with major job boards hosting millions of active listings. The migration of job matching onto the internet was anticipated to lower search costs and improve worker–firm matching, though early analyses also cautioned that its benefits would be uneven and would generate new frictions (Autor, 2001). A structural feature of these platforms is that posting a job has become inexpensive and frictionless, while the act of applying—tailoring a resume, composing a cover letter, researching the employer, completing application forms—remains effortful for the applicant. This asymmetry in cost has a consequence that is easy to overlook: it changes who does the work of screening.

In a traditional hiring process, much of the screening burden sat with the employer. A posting generally signaled a genuine, immediate opening, and a human reviewer assessed applicants. Under digital hiring, two shifts complicate this picture. First, because postings are cheap to create and maintain, a listing no longer reliably signals an immediate intent to hire; it may represent talent-pipeline building, an administrative artifact, a posting that is no longer current, or, in some cases, outright fraud (Federal Trade Commission, 2025; Ng, 2024); recent empirical work has begun to quantify the prevalence and labor-market costs of such non-genuine postings (Singer and Oktay, 2025). Second, automated systems increasingly mediate the employer side, with applicant tracking systems (ATS) filtering candidates before human review (Fuller et al., 2021). The combined effect is that, before an application is worth making, the seeker must now perform a prior task: screening the posting itself—inferring whether it represents a real opportunity, whether automated filtering will exclude them regardless of fit, and whether the role is even viable given their profile.

This Perspective argues that the digitalization and partial automation of hiring has reallocated a layer of screening labor from employer to job seeker, and that this shifted burden is uncompensated and largely unrecognized. We do not claim that this burden is new in kind, nor that it falls on a precise share of postings; our aim is conceptual. We offer a way of naming and organizing the forms of opacity that a job seeker must now screen for, anchored in the economics of information asymmetry (Akerlof, 1970), and we use observations from a job-analysis platform to illustrate that these forms occur routinely in real job-seeking activity.

The phenomenon connects to a broader literature. Prolonged, low-feedback job search has long been associated with stress and discouragement (Paul and Moser, 2009; Wanberg, 2012), and labor-market frictions are understood to produce inefficient matching (Marinescu and Rathelot, 2018). If a meaningful portion of search effort is consumed by screening that the platform environment has displaced onto the applicant, this represents a transformation in the everyday experience of work-seeking—a topic squarely within the study of how digitalization and the automation of jobs reshape society. These dynamics are relevant to the UN Sustainable Development Goals (United Nations, 2015), particularly SDG 8 (Decent Work and Economic Growth) and SDG 9 (Industry, Innovation, and Infrastructure), insofar as the efficiency and transparency of digital labor-market infrastructure bear on the quality of work-seeking.

2 The shifted burden: an information-asymmetry account

The mechanism underlying the shifted burden of search is information asymmetry. In Akerlof’s classic account of markets with quality uncertainty, sellers possess information about the true quality of a good that buyers cannot directly observe, and this gap degrades the functioning of the market (Akerlof, 1970). A complementary tradition examines how parties attempt to resolve such asymmetry through costly signals, as in Spence’s analysis of the labor market, where applicants invest in observable signals to convey unobservable quality to employers (Spence, 1973). The shifted-burden account turns attention to the other side of this relationship: the effort a job seeker must expend to read the employer’s signals—to infer, from the posting alone, what the employer already knows. Digital hiring exhibits a structurally similar asymmetry. The posting entity knows whether a listing reflects an immediate intent to hire, whether it is oriented toward building a candidate pool, whether a preferred internal candidate already exists, or whether the listing is no longer current. The job seeker observes only the text of the posting and cannot directly verify any of this.

Crucially, the seeker cannot costlessly resolve this uncertainty. Determining whether a posting is genuine, whether its requirements signal heavy automated filtering, or whether the role is viable requires effort—reading closely, cross-referencing, and judging. That effort is the shifted burden. It is incurred before, and independently of, the effort of applying, and it is incurred by the party with the least information. Where a traditional process placed screening with the better-informed employer, the digital environment relocates a portion of it to the less-informed applicant.

This account reframes several familiar complaints about modern job search—“ghost jobs,” opaque requirements, automated rejections—not as isolated grievances but as surface expressions of a single underlying shift in who bears screening labor. It also clarifies what a remedy would require: reducing the information asymmetry at its source, rather than asking job seekers to become ever more sophisticated screeners.

3 A taxonomy of hiring-intent signals

To make the shifted burden concrete, it helps to name the specific forms of opacity a job seeker must screen for. Drawing on a job-analysis platform developed by the author, we organize observable posting characteristics into a taxonomy of hiring-intent signals (Figure 1). The taxonomy is offered as a conceptual and descriptive device—a vocabulary for the kinds of uncertainty a seeker confronts—and not as a validated diagnostic instrument; it has not been checked against employer-confirmed ground truth, a limitation we return to below.

The categories are as follows. High Intent postings carry clear signals of a genuine, immediate opening. High ATS Filtering Risk postings present rigid, checklist-style requirement structures suggesting heavy reliance on automated keyword screening that may exclude otherwise-qualified candidates before human review (Fuller et al., 2021). Low Transparency/Pipelining postings appear oriented toward building a candidate pool rather than filling an immediate vacancy. Low-Effort/Vague postings provide too little detail for a seeker to assess the opportunity. Internal Hire Likely postings carry signals suggesting a predetermined candidate. High Scam Risk postings exhibit indicators consistent with fraudulent intent (Federal Trade Commission, 2025). Each category corresponds to a distinct screening question the seeker must answer at their own cost.

Beyond hiring intent, a related screening task concerns viability: even a genuine posting may be a poor match on a dimension that effectively precludes success, such as a required core skill the applicant lacks, an incompatible work location or arrangement, or stated compensation outside the applicant’s range. We refer to such factors as deal breakers. They represent a second form of pre-application screening labor—assessing fit against the detailed requirements rather than the title or summary—that, when neglected, leads to effort spent on applications unlikely to succeed.

4 Illustrative observations

As a limited illustration—and no more—we draw on observations from 784 job-posting analyses conducted through the author’s platform between October 2025 and May 2026, solely to show that the screening questions named above arise in practice rather than only as theoretical possibilities. These observations are descriptive of this particular set of analyses only. They are not a sample designed to estimate how common any signal is in the labor market, and they should not be read that way: the analyses are heavily clustered, with the large majority contributed by a small number of users, predominantly based in the United States and concentrated in software-related roles (see Limitations). With that strict caveat, the observations are useful for one narrow purpose: showing that each category in the taxonomy corresponds to postings that genuinely occur. These observations carry no analytical weight in the argument that follows; they serve only to confirm that the forms of opacity named in the taxonomy are encountered in practice, and the conceptual claims of this Perspective stand independently of them.

Across the 784 analyses, every posting received a hiring-intent classification (Table 1). Postings carrying a rigid, checklist-style requirement structure—our High ATS Filtering Risk signal—were the most frequently observed non–High-Intent category in this set. Pipelining-style and low-effort postings appeared in smaller numbers, and internal-hire and scam signals were rare within these particular analyses. The point of the table is not the specific proportions, which reflect this clustered set of users and roles, but the simple fact that each form of opacity named in the taxonomy was instantiated by real postings that seekers had encountered and chosen to examine.

Table 1. Hiring-intent signals observed among 784 analyses conducted through the author’s platform.

Counts describe this particular, clustered set of analyses and are not estimates of prevalence in the labor market.

The viability-screening task was likewise instantiated. Among these analyses, a substantial share of postings—300 of the 784 (38.3%)—carried at least one deal breaker, most commonly a core-skill mismatch, in which the detailed requirements revealed an incompatibility not apparent from the title or summary. This illustrates the second layer of the shifted burden directly: a seeker who relied on the title alone would have invested application effort, whereas a closer reading—itself a form of labor—would have revealed the mismatch in advance. Again, the value of the observation is illustrative; it demonstrates that viability screening is a real and recurring task, not that mismatches occur at any particular rate in the wider market.

Taken together, these observations support a modest but, we think, useful claim: the screening questions implied by the shifted-burden account are not hypothetical. Real job seekers, examining real postings, encounter opacity of exactly the kinds the taxonomy names, and they expend effort resolving it.

5 Discussion

5.1 Screening labor as a feature of digital hiring

The central implication of this Perspective is that a portion of the labor of hiring has been displaced onto the party least equipped to perform it. This reframing matters because it shifts attention from individual coping strategies—“how can a job seeker spot a ghost job?”—to the structural question of why the burden falls on the seeker at all. The information asymmetry is a property of the platform environment, not a deficiency of individual applicants. Treating it as the latter risks normalizing an ever-growing expectation that job seekers become expert screeners, when the more consequential lever lies in reducing the asymmetry at its source.

5.2 Connections to human-centered and human-in-the-loop design

The shifted-burden framing aligns with a broader concern in human-centered computing: that automated systems should be designed around the people who must live with their outputs, with appropriate human oversight and transparency, and in a manner attentive to the values of those they affect (Gabriel, 2020). Recent work on human-in-the-loop and human-centered AI in high-stakes settings has emphasized exactly this balance between automation and human judgment (Raghavan et al., 2020), and studies of applicant reactions show that job seekers themselves perceive algorithm-driven selection as less fair than human or human-assisted processes (Lavanchy et al., 2023). Applied to hiring, the same principle suggests that platforms could be designed to surface, rather than obscure, the information seekers currently expend effort inferring—for example, by making posting status, recency, or screening practices more transparent.

5.3 Design and policy directions

If opacity is the source of the shifted burden, then transparency is the natural direction of remedy. Platform-level interventions could include clearer indication of whether a posting reflects an immediate opening, how long it has been active, and whether automated screening is applied. Research on algorithmic hiring tools has documented how opaque automated screening can affect outcomes in ways that are difficult to audit (Raghavan et al., 2020), how such tools can reproduce demographic biases (Armstrong et al., 2024), and how their use raises distinct ethical questions around transparency and fairness (Hunkenschroer and Kriebitz, 2022; Hunkenschroer and Luetge, 2022)—together underscoring the value of transparency as a design goal. At the policy level, several jurisdictions have begun to consider job-posting transparency measures; the shifted-burden account offers one rationale for such measures, namely that they would return screening information to the party currently forced to infer it. A concrete precedent exists in pay-transparency legislation: a growing number of U.S. states now require employers to disclose salary ranges directly in job postings, compelling disclosure of information that applicants previously had to infer or extract through costly back-and-forth. The shifted-burden account suggests the same regulatory logic could be extended beyond compensation to other currently-opaque signals—whether a posting reflects an active opening, how long it has been live, and whether automated screening is applied—each of which currently forces the seeker to expend screening effort that disclosure would eliminate. We advance these as directions for consideration rather than as evaluated recommendations.

6 Research propositions and future directions

The shifted-burden account is a conceptual framing, and its value ultimately depends on whether it can be investigated empirically. We offer the following propositions as starting points for systematic inquiry rather than as established results. Each is stated so that it could, in principle, be tested with an appropriate study design.

Proposition 1The burden is measurable. The time a job seeker spends evaluating a posting before deciding whether to apply—assessing genuineness, filtering risk, and viability—is a distinct and measurable component of search effort, separable from the time spent preparing applications. This could be tested through time-use diaries or instrumented observation of the pre-application phase.

Proposition 2Opacity raises the burden. The less a posting reveals about its true hiring intent, the more screening effort a seeker must expend before reaching a confident apply/skip decision. This predicts a measurable relationship between posting opacity and per-posting evaluation time, testable by varying or coding opacity and measuring decision effort.

Proposition 3Transparency lowers the burden. Surfacing otherwise-hidden information—whether a posting is currently active, how long it has been open, or whether automated screening is applied—reduces the screening effort seekers expend. This is testable through controlled comparison of postings shown with and without such disclosures.

Proposition 4The burden may fall unevenly—open question. Whether the shifted burden is distributed unequally across job seekers—for example, by financial resources, experience, or access to professional networks that provide alternative verification channels—is an open distributional question. The present observations cannot address it, and we explicitly do not claim such an effect; we flag it as the question most in need of dedicated, representative study.

Pursuing these questions would require study designs quite different from the illustrative observations used here: representative or randomized sampling across platforms, larger and more diverse participant pools, direct measurement of seeker effort and time, and—for the taxonomy itself—validation against employer-confirmed hiring outcomes. We regard such validation as a priority, since the taxonomy is presently a descriptive vocabulary rather than a verified instrument.

7 Limitations

This is a Perspective, and its empirical content is strictly illustrative. Several limitations bound any reading of it.

Clustered, non-representative observations. The 784 analyses were contributed through a single platform, and the large majority were contributed by a small number of highly active users (approximately 90% by two to three users). The analyses are predominantly U.S.-based and concentrated in software-related, senior roles. They therefore cannot support any claim about how common a signal is in the labor market, and we make none; the counts reported describe only this clustered set.

User-selected postings. Users chose which postings to analyze rather than sampling randomly, which may bias the set toward postings they already found notable or were genuinely considering.

Unvalidated taxonomy. The hiring-intent categories reflect observable posting characteristics, not employer-confirmed intent. Without ground-truth validation, the taxonomy should be treated as a descriptive framework, not a diagnostic test.

Author-operated platform. The data were generated through a commercial platform developed and operated by the author, and the hiring-intent categories—including the High ATS Filtering Risk label—were produced by that platform’s own classification logic. The observations therefore illustrate the platform’s taxonomy applied to postings its users chose to analyze; they are not an independent verification of that taxonomy. This is a substantial conflict of interest, and we have accordingly confined the data to a strictly illustrative role, with conservative interpretation throughout.

These constraints are the reason the paper’s claims are deliberately modest: a conceptual reframing of where screening labor falls, supported by observations that the named forms of opacity genuinely occur, and nothing more.

8 Conclusion

Digital platforms have made job postings cheap to create while leaving applications costly to make, and automated screening has further reshaped the employer side of hiring. A quiet consequence is that a layer of screening labor—judging whether a posting is genuine, whether automated filtering will exclude an applicant, and whether a role is viable—has shifted from the employer to the job seeker. This Perspective has named that shift, grounded it in the economics of information asymmetry, and offered a taxonomy of the forms of opacity seekers must now screen for, illustrated by observations that these forms occur routinely in practice. We have been careful to claim no more than the evidence allows: the observations are clustered and illustrative, not representative. The contribution is conceptual—a lens that moves the question from how job seekers can cope with opacity to why they are made to bear it, and toward platform and policy designs that would return that information to the better-informed party. We hope the framing and the propositions offered here are useful starting points for empirical work on the everyday labor of job search in digital markets.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI was used for language editing and drafting assistance; all conceptual content, analysis, and interpretation are the author’s own.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Sandro Serpa, University of the Azores, Portugal

Reviewed by: Veera Venkata Raghunath Indugu, Bellevue University, United States