Executive Summary
OpenAI has released a research report detailing how American workers are using ChatGPT to close the wage information gap, sending nearly 3 million compensation-related messages daily. The analysis shows that users most often seek help translating pay into usable benchmarks, especially in fields with less transparent or more variable pay structures. Alongside the report, OpenAI introduced WorkerBench, a new benchmark for evaluating AI accuracy on labor market tasks, noting that its GPT-5.4 model shows high accuracy in initial tests.
Key Takeaways
* Usage Volume: In the US, users send an average of nearly 3 million messages per day to ChatGPT asking about wages, compensation, or earnings.
* Common Queries: The most common queries are for pay calculation (26%), specific role compensation (19%), and entrepreneurial income potential (18%).
* Demand Pattern: Searches are most concentrated in higher-skill fields where pay is less transparent and more negotiable, such as creative arts, management, healthcare, and tech.
* New Evaluation Benchmark: OpenAI introduced "WorkerBench," a new initiative to evaluate how well AI models perform on tasks valuable to workers, starting with wage benchmarking.
* Model Performance: In an initial evaluation using WorkerBench, the GPT-5.4 model was found to be highly accurate with low bias when its wage estimates were compared against 2024 OEWS median wage data.
Strategic Importance
This announcement positions ChatGPT as a critical tool for labor market transparency, showcasing a high-value, practical use case that empowers individual workers while demonstrating a commitment to improving model reliability in this economically significant domain.