Finding a job is rarely straightforward. While economists have long studied job search frictions, the rise of online job search platforms has transformed the landscape. Yet, despite the wealth of information available, many jobseekers still struggle to find suitable employment. A pressing policy question is whether digital tools – particularly occupation recommendation systems – can improve job search outcomes.
Governments and employment services have increasingly turned to technology to assist jobseekers. Automated occupation recommendations promise to help workers navigate a complex job market, identify better job matches, and potentially shorten unemployment spells. But do they work? In this column, we review existing research and presents new evidence from a recent randomised controlled trial on the effectiveness of occupation recommendations.
Several economic mechanisms suggest that occupation recommendations could be beneficial for jobseekers due to their limited knowledge of job opportunities, occupational mismatch, and the potential for personalised job search strategies. Most jobseekers do not have complete knowledge of all vacancies that might suit them. Instead, they tend to search narrowly, often focusing on jobs similar to their previous employment. This is a rational strategy given the complexity of the job market, but it can lead to missed opportunities.
Prior research has highlighted the limitations of jobseeker search strategies. Belot et al. (2019) found that jobseekers often underestimate the breadth of opportunities available to them. Their study in the UK demonstrated that providing jobseekers with additional occupational information led to more applications and better job-finding outcomes. Similarly, Altmann et al. (2022) conducted an experiment in Denmark showing that online job search tools that introduce jobseekers to new opportunities can significantly impact search behaviour and outcomes.
Occupation recommendation systems aim to address the gap by systematically broadening the set of job opportunities considered by jobseekers (Le Barbanchon et al. 2023, Ben Dhia et al. 2022). Recommender systems use data on past job mobility patterns, skill transferability, and labour market conditions to generate job suggestions that a worker might not have otherwise considered.
Labour market mismatch is pervasive – many workers are employed in jobs that do not fully utilise their skills, while others struggle to transition to roles better suited to their abilities (Sahin et al. 2014). Mismatch can lead to lower wages, job dissatisfaction, and higher turnover rates. A key challenge is that many jobseekers base their search on their previous job rather than their underlying skills. However, prior jobs may not always reflect a worker’s full capabilities or optimal career trajectory. For instance, someone who previously worked as a retail cashier may have strong analytical skills but might never consider roles in data analysis or logistics.
Our intervention was designed to compare the effectiveness of two different types of occupation recommendations (Bächli et al. 2025). The experience-based approach provided jobseekers with suggestions that closely aligned with their previous job, leveraging their work history to identify potential matches. This method assumes that jobseekers benefit from remaining in familiar occupational fields where they have accumulated specific human capital.
The profile-based approach, by contrast, recommended occupations based on a jobseeker’s assessed skill set, rather than past employment. This method allowed for a more flexible job search, particularly for individuals whose prior jobs did not align well with their underlying abilities. By comparing these two approaches in a randomized controlled trial, we aimed to evaluate which method was more effective in improving employment outcomes.
The intervention was implemented through an online job search platform, Jobs for You (J4U), developed specifically for this study. Participants were recruited in collaboration with public employment services in Switzerland and were randomly assigned to one of three groups: (1) experience-based recommendations, (2) profile-based recommendations, or (3) a control group that had access to the platform but without occupation recommendations.
Jobseekers in the treatment groups received weekly updates with tailored job suggestions, while the platform tracked their search behaviours, including login activity and job application clicks. The intervention lasted for eight months, during which we measured job-finding rates, the quality of jobs obtained, and whether jobseekers moved into occupations better aligned with their skills. Our collaboration with the Swiss Federal Statistical Office allowed us to link jobseekers’ search behaviours with administrative employment records, providing robust data on employment outcomes.
The results show that jobseekers with profile-based recommendations looked at job advertisements that were better matched with their skills. Profile-based recommendations are particularly effective for jobseekers who were previously mismatched in their roles. These jobseekers found employment faster when guided toward jobs aligned with their skills rather than their past job titles (Figure 1). Personalised, profile-based recommendations were especially beneficial for jobseekers with limited work experience and a history of job mismatch. For these individuals, shifting away from their previous occupations and toward roles aligned with their skills led to faster job placement.
Figure 1 Job-finding rates by treatment groups
Notes: This figure shows the proportion of jobseekers who found a job across three groups: experience-based = received recommendations with respect to previous occupation, profile-based = received recommendations with respect to skills, and controls = no recommendations.
One of the most promising aspects of occupation recommendations is their ability to be personalised. Unlike traditional job counselling, which often provides generic advice, digital platforms can tailor recommendations to individual jobseekers based on their skills, in addition to matching their work history and preferences. This skill-based personalisation is particularly important for workers facing major career transitions. Consider an individual who worked for years in an industry experiencing structural decline. A system that recognises their transferable skills and suggests alternative occupations with strong demand can significantly improve their chances of re-employment.
Automated occupation recommendations represent a promising tool for improving job search efficiency and addressing labour market mismatch. Our research, along with previous studies by Belot et al. (2019) and Altmann et al. (2022), suggests that these recommendations can help jobseekers explore broader opportunities, find jobs faster, and achieve better job matches.
However, the effectiveness of these recommendations depends on their design. A well-calibrated system that personalises suggestions based on an individual’s skills, rather than just past work experience, can make a significant difference in employment outcomes.
Given the potential benefits of occupation recommendations, policymakers should consider integrating them more systematically into public employment services. For instance, providing jobseekers with occupation recommendations that extend beyond their immediate work history can improve job-finding outcomes. Effective occupation recommendations require accurate data on worker skills. Public employment agencies should invest in better assessments to facilitate personalised recommendations. While experience-based recommendations may work well for some jobseekers, profile-based recommendations can be more effective for those facing job mismatch.
As governments and employment agencies continue to digitise job-matching services, ensuring that these systems effectively serve diverse jobseekers will be crucial. Future research should further explore how occupation recommendations can be optimised to improve labour market efficiency and economic mobility.
References
Altmann, S, A M Glenny, R Mahlstedt, and A Sebald (2022), “The direct and indirect effects of online job search advice”, IZA Discussion Paper 15830.
Bächli, M, R Lalive, and M Pellizzari (2025), “Automated and personalized job recommendations for jobseekers: Evidence from a randomized intervention”, CEPR Discussion Paper 19949.
Le Barbanchon, T, L Hensvik, and R Rathelot (2023), “How can AI improve search and matching? Evidence from 59 million personalized job recommendations”, Uppsala University.
Belot, M, P Kircher, and P Muller (2019), “Providing advice to job seekers at low cost: An experimental study on online advice”, Review of Economic Studies 86(4): 1411–47.
Ben Dhia, A, B Crépon, E Mbih, L Paul-Delvaux, B Picard, and V Pons (2022), “Can a website bring unemployment down? Experimental evidence from France”, NBER Working Paper 29914.
Sahin, A, J Song, G Topa, and G L Violante (2014), “Mismatch unemployment”, American Economic Review 104(11): 3529–64.
