Loading page…
Research covering 25,000 workers finds task changes and moves into better-paid occupations, but no measurable earnings or employment effects through December 2024.
Listen to this story
Rather than producing an immediate broad wage dividend, chatbot adoption is adding coordination, review and AI-related work across exposed jobs—including for some workers who never use the tools themselves. Brookings researchers found no measurable average change in earnings or hours through 2024, while adopters were more likely to move into better-paying occupations where chatbot skills mattered. The evidence is an early-adoption snapshot and does not establish what labor-market effects look like in 2026.
Around 4% of non-users reported new AI-related workloads; among teachers who had never used chatbots, the figure was 10%, mostly due to students’ AI use.
New tasks involved generating content or insights roughly 40% of the time, overseeing AI about a third, and integrating tools about a quarter.
Although 43% of workers said their employers encouraged chatbot use in 2024, 6% faced bans; adoption reached 93% where encouragement, enterprise tools and training were combined.
AI chatbots are changing what Danish workers do without measurably changing their pay or hours, according to research published by Brookings on October 6, 2026. The study tracks 25,000 workers in highly exposed occupations. It finds new work around AI—even for non-users—but no measurable earnings benefit from adoption during the period studied.
Anders Humlum and Emilie Vestergaard linked survey responses across 11 occupations and about 7,000 workplaces to monthly administrative records through December 2024. The occupations included software development, marketing, law and teaching. They compared chatbot users with similar non-users in the same occupation, and adopting workplaces with similar workplaces without chatbot initiatives.
Chatbot users earned more than comparable non-users, but those gaps existed before ChatGPT launched. Tracking the groups over time revealed no differential earnings or hours changes; the estimates ruled out average effects larger than 2%. The null findings also held for daily users and workers reporting savings of more than an hour a day.
Workplaces encouraging chatbot use showed no differential changes in headcounts, wage bills, job creation or destruction, or the mix of hires and departures. Those results concern an early adoption window ending in 2024, not labor-market outcomes in 2026.
About 4% of workers who had never used chatbots reported new workloads resulting from the tools.
Among teachers who had never used chatbots, 10% reported new AI-related work, mostly responding to students’ AI use.
In the 2024 survey, 43% of workers had employers that explicitly encouraged chatbot use, while 6% faced bans. Adoption reached 93% where workplaces combined encouragement, enterprise tools and training. Most users—85%—said they spent the time saved on other job tasks, rather than simply doing more of the same work or taking more leisure.
One labor-market outcome did stand out: chatbot adopters were more likely to switch occupations, moving into better-paying roles where the tools were more relevant. The authors interpret this as consistent with AI helping workers access otherwise scarce expertise. That interpretation is distinct from finding an average earnings lift across chatbot users.
The study also addresses concerns about entry-level jobs. Early-career employment had fallen in AI-exposed occupations, but the researchers found that firms adopting generative AI were not driving those declines. Their finding cautions against treating a job’s potential exposure to AI as evidence that actual adoption caused its employment losses.
Loading discussion...
Join the conversation
Explain how you would divide the benefits between workers and employers.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.