AI Freezes Entry-Level Hiring More Than It Cuts Jobs, Study Finds

Artificial intelligence is reshaping the global labor market not by triggering mass layoffs, but by quietly closing the door on entry-level hires — freezing the traditional on-ramps that young workers depend on to launch their careers. That is the central finding of new research from Peking University, presented at the 2026 Summer Davos forum in Dalian.

Professor Zhang Dandan of Peking University’s National School of Development, speaking in an interview with Caixin, explained that macro employment and unemployment rates remain stable because companies prefer to freeze or reduce external recruitment rather than proactively firing existing staff. “Even when small-scale layoffs occurred,” Zhang said, “many affected workers could find new jobs quickly, preventing an extension of an unemployment cycle.”

The Research Behind the Finding

Zhang and her collaborator, Professor Li Jia of Singapore Management University, analyzed approximately 1.63 million online job postings from Chinese recruitment platform Zhaopin (智联招聘) to construct an AI-LLM Occupational Exposure Index. The index reveals a troubling pattern: white-collar positions are disproportionately concentrated in occupations with high AI exposure risk, and the higher that risk, the more employers raise entry barriers for junior applicants.

Cognitive jobs — including copywriting, customer service, junior programming, drafting legal documents, translation, and research report writing — are simultaneously entering the scope of AI substitutability. As CGTN reported from the Summer Davos forum, the question of whether AI will replace entry-level jobs has become a central concern for policymakers and business leaders alike.

This finding is corroborated by international research. A 2025 Harvard paper found that AI-adopting firms cut entry-level hiring and spare senior workers through slower hiring, not layoffs. According to Anthropic’s March 2026 report, AI theoretically handles 94% of computing and math tasks but is actually used for only 33% — with 57% of usage augmenting humans and 43% replacing them.

Why This Wave of AI Is Different

The researchers argue that this round of AI disruption is historically unprecedented in three ways. First, its breadth: AI targets cognitive jobs across multiple industries simultaneously — not confined to a single sector like manufacturing. Second, its speed: while electricity took roughly 30 years to permeate industry and the internet took about 10, large language models can be deployed globally via cloud computing instantly. Third, its horizontal diffusion: technological disruption is unfolding synchronously across industries and occupational tiers.

As Li and Zhang wrote in a Caixin opinion piece, “When the pace of AI substitution systematically outstrips the economic system’s pace of creating new jobs and workers’ pace of retraining and occupational transition, the problems brought by technological progress are no longer merely short-term frictional unemployment, but may evolve into persistent aggregate demand shortfalls and social welfare losses.”

The “Aggregate Demand Externality” Problem

A key insight from the research is what Li and Zhang call the “aggregate demand externality” problem. Each firm internalizes the cost savings from automation but does not account for the consequence that displaced workers, with reduced incomes, consume less — eventually weakening demand across the entire economy. This is treated not just as a job-loss problem, but as a market failure requiring government intervention.

As they wrote, “The cost-saving benefits from corporate automation are typically captured exclusively by the firms themselves, while the demand decline triggered by employment contraction is collectively borne by society as a whole.”

A Proposed Solution: The AI Employment Compensation Fund

To address this challenge, Li and Zhang have proposed a “one fund, two pillars, three supporting measures” framework. The centerpiece is an AI Employment Compensation Fund financed through fiscal allocations, unemployment insurance surpluses, employer contributions, and potential surcharges on data and computing power revenues.

The two pillars divide the intervention timeline: in the short term, an AI-exposure-based employment risk monitoring and early warning system; in the long term, institutionalized lifelong learning through portable “skills accounts” and “learning accounts” that move with workers across jobs. The three supporting measures include adapting social security for flexible employment, expanding unemployment insurance’s role in skills transitions, and channeling productivity dividends into pension systems.

As Fred Gao’s Inside China newsletter summarized, the proposal reflects a broader shift in Chinese scholarly thinking: technology does not determine social outcomes on its own. Whether AI broadens shared prosperity or deepens social divisions depends on how governments mediate the relationship between technology and labor.

What to Watch For

The research raises urgent questions for policymakers worldwide. Will China implement the proposed AI Employment Compensation Fund? Can education and training systems adapt quickly enough to prepare workers for an AI-augmented economy? And perhaps most critically: will the freezing of entry-level hiring become permanent, or will new types of entry-level roles emerge?

For young workers entering the job market, the message is clear: the traditional pathway of starting at the bottom and learning on the job is being disrupted. As Singapore’s Prime Minister Lawrence Wong put it in a speech cited by the researchers, “We may not be able to save every job, but we must protect every worker.” The challenge now is building the institutions to make that protection real.