Impact of ChatGPT and Large Language Models on the U.S. Labor Market

OpenAI released a research report analyzing how GPT‑4 and related large language models (LLMs) could reshape the U.S. labor market. By mapping occupations from the O*NET 27.2 dataset (1,016 jobs) to detailed work activities (DWAs) and using human annotators together with GPT‑4, the study measured the probability that a GPT‑driven system would cut the time required for a task by at least 50%.

The authors introduced an "exposure rubric" (E1, E2) to quantify this impact and derived three aggregate metrics: α (E1 alone), β (E1 + 0.5·E2), and ζ (E1 + E2). Results show average occupation‑level α values around 0.14‑0.15 (≈15% of tasks directly exposed), β values exceeding 30%, and ζ values surpassing 50%.

Consequently, the analysis predicts that about 80% of the U.S. workforce will have at least 10% of their tasks influenced by GPT, while roughly 19% of workers could see more than half of their tasks affected. High‑income jobs, which often involve software‑centric activities, face greater risk, whereas physically intensive sectors such as food production, forestry, and social assistance show minimal exposure.

Industry‑level breakdowns identify data‑processing services and publishing as the most vulnerable, while the study also highlights that occupations requiring scientific reasoning and critical thinking are negatively correlated with LLM exposure, whereas programming and writing skills show a positive correlation.

Figures in the original report (included as images) illustrate overall economic impact, wage‑exposure relationships, and differences across O*NET job zones, confirming that higher‑skill, higher‑education roles tend to experience stronger LLM influence.

The authors acknowledge methodological limitations, such as the bias inherent in labeling tasks and the current propensity of LLMs to generate hallucinations, which necessitates continued human oversight. Nonetheless, they argue that widespread LLM adoption will reshape work, potentially displacing many jobs while also creating opportunities for higher‑quality employment.