AI paradox in job market: junior professionals lose out, experienced gain

Artificial intelligence is changing the job market, but so far has not led to the feared widespread wave of layoffs. Instead, a more differentiated picture is emerging, one that is particularly alarming for young people: Generative AI is primarily affecting career starters.

An updated study by the Stanford Digital Economy Lab, led by economists Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen, uses data up to June 2026 to show that the gap between experienced employees and junior staff in AI-exposed professions has widened further. Established professionals are reportedly benefiting from the technology. Meanwhile, opportunities are increasingly being blocked for junior staff.

The investigation is based on a comparatively robust dataset. Instead of surveys or job advertisements, the researchers analyzed the payroll data of the US personnel service provider ADP. This dataset covers between 3.5 and five million employees monthly from January 2021 to mid-2026. The authors determine the AI susceptibility of professions using a classification largely developed by OpenAI researchers and the Anthropic Economic Index, which is based on real usage data.

Gap for career starters is widening

The numbers initially show a positive development: overall employment increased by six percent during the study period. Even in the professions most affected by AI, there was still a four percent increase. Therefore, there can be no talk of widespread job cuts. The real effect only becomes apparent when looking at age groups: among 22- to 25-year-olds working in the professional fields most threatened by AI, employment has decreased by about eleven percent since the launch of ChatGPT in November 2022. In contrast, employment in professions within the same age group with low AI exposure has increased by about ten percent.

This creates a gap of 19 percent, which was still 15 percent in the summer of 2025. The fact that this gap is widening makes the results robust, according to the scientists. This is by no means a temporary dip due to the economic situation, as they have already considered factors such as interest rate development, education level, or home office opportunities in their analysis.

The decline rarely occurs through dismissals. Those already within a company retain their jobs with a similar probability as before. The pressure to adapt is instead released through a gradual reluctance in new hires. Companies are simply no longer filling junior positions. There are no spectacular layoff waves, but at most jobs that quietly disappear.

The identified effect is also concentrated on professions where AI automates human tasks. Where it acts as a supplement, employment remains stable or increases. Experienced employees use AI complementarily based on their existing knowledge. Junior staff, whose work often consists of standardized data preparation and text drafts, can be more easily replaced. Wages hardly react to this shift, and thus the market transfers the pressure to new entrants.

A global phenomenon from Asia to Europe

That this is not purely a US phenomenon is shown by international data, according to the FAZ. An analysis by the Bank of Korea reveals a strong polarization: of nearly 295,000 lost jobs among 15- to 29-year-olds, 94 percent were in AI-intensive industries. In parallel, employment of 50- to 59-year-olds in these sectors has reportedly increased significantly. In Germany, according to an Ifo survey from June, around half of companies expect declining entry-level wages for junior staff without academic degrees due to AI.

Supplementary data from the International Labour Organization indicates that global youth unemployment rose to 12.4 percent in 2025. Mid-skilled entry-level positions are particularly at risk. Furthermore, OpenAI data from US companies shows that, ironically, career starters use ChatGPT most intensively. As their tasks are highly standardized, they unintentionally make themselves replaceable through its use. Companies are cutting jobs where the professionals of tomorrow would be trained, which endangers the expertise pipeline in the long term. Pure AI skills like prompting are not sufficient to compensate for professional gaps, according to other studies. Therefore, sound professional knowledge remains indispensable.

(vbr)