New Cedefop analysis of online job advertisements across EU Member States suggests that the rise of generative AI is accompanied by a notable shift in the composition of labour demand with growing signs of renewed interest in vocational skills.
A shifting vacancy landscape
The European labour market is evolving. Cedefop's analysis of Eurostat online job advertisement (OJA) data points to a gradual but visible recomposition of vacancy demand since the emergence of generative AI tools in late 2022.
Occupations with high exposure to AI capabilities are seeing their share of total job postings decline. Software developers, sales and marketing professionals, client information workers, and database specialists are among the occupations that have lost the most ground in relative terms. At the same time, occupations that rely on physical presence, manual expertise, and applied technical skills, for example, engineering technicians, machinery mechanics, construction trades, and transport workers have seen their share of vacancies grow.
These patterns are observed across most of the EU Member States. The analysis is based on vacancy shares rather than absolute counts, making it robust to known measurement characteristics of online job advertisement data.
Figure 1. Vacancy share change by occupation and AI exposure
Vocational occupations: a trend reversal
A broader pattern emerges when looking at vocational occupations as a group. Between 2019 and mid-2022, the share of EU online vacancies in VET-related occupations had been declining steadily, roughly, from above 36% to below 33%. This trend appeared to reflect the growing weight of knowledge-intensive and digital occupations in the labour market.
Since late 2022, however, the trend has reversed. The VET share of vacancies has recovered gradually, returning to above 36% by early 2025. While multiple factors may be at play, such as the post-pandemic recovery in construction and manufacturing, the timing of the reversal coincides with the broad adoption of generative AI tools and a corresponding contraction in demand for several highly AI-exposed, non-vocational occupations.
Figure 2. Share of vacancies in VET occupations in the EU
Implications for skills policy
The long-term durability of these patterns is yet to be confirmed. However, these early signals deserve attention within the context of EU’s skills policy agenda. They suggest that vocational education and training systems equip workers with capabilities that retain their value in a changing technological environment. This is particularly relevant for occupations involving hands-on expertise and in-person service, which are not easily replicable by AI.
Cedefop will continue to monitor these developments as new data becomes available, contributing to the evidence base for Europe's skills and employment policies.