Dallas Federal Reserve finds link between generative AI exposure and fewer job ads in Texas

methodology behind the numbers

The report released in recent days relies on mapping O*NET tasks, the U.S. Department of Labor’s occupational database, to actual usage patterns in Anthropic’s Claude model. Researchers cross-referenced the resulting scores with millions of job ads from Lightcast, a labor-market data platform. The result: the more a profession is “exposed” to the model’s capabilities, the fewer postings companies in Texas publish for that profession.

gap opened with ChatGPT

At the occupational level, the gap began to appear immediately after the release of ChatGPT at the end of 2022. By the end of 2023 the gap measured about 5% fewer ads in exposed professions compared with less exposed ones, and it widened to about 8% in the first quarter of 2025. At the firm level, companies that already employed workers in high-exposure professions recorded an 8% to 9% decline in new postings by the beginning of 2026.

who is affected first

Because most online postings require little experience, the decline hits recent graduates and career switchers before it reaches senior employees. The professions with the highest exposure scores include software development, web design, management, clerical work and editing—areas where writing, coding and information organization are core daily tasks.

measurement based on actual use, not forecasts

Choosing Claude as the exposure benchmark is notable in itself: rather than relying on intention surveys or theoretical benchmarks, the researchers used real-world usage data to determine which tasks the model already performs today. This brings the metric closer to field reality, but also limits it to the current capabilities of a single specific model.

question mark about the future

The report does not establish definitive causality; part of the decline may stem from other macro-economic factors that affected the same sectors simultaneously. Nevertheless, the timing closely following ChatGPT’s launch and the consistency between occupational and firm-level effects strengthen the hypothesis that generative AI is already reshaping demand for junior workers in text- and code-intensive professions.