AI Language Models Leave Distinct Mark on Economics Research Writing, Study Shows

How AI Language Models Are Changing Economics Research Writing

A recent study from the University of Massachusetts Amherst reveals that large language models (LLMs), like ChatGPT, are increasingly influencing how economists write their research papers. By examining writing styles in 25 leading economics journals over 24 years, the study shows a clear shift in language patterns since ChatGPT's introduction in late 2022.

The researchers found that AI-related language features rose nearly 5 percentage points in papers published during 2023–24. This increase more than doubled from about 3 points in 2023 to nearly 7 points in 2024. This suggests economists are growing more comfortable using LLM tools to assist in their writing.

Identifying AI’s Linguistic Fingerprint

The study used a novel approach by tracking the frequency of 25 AI-associated words—terms like “underscore,” “nuance,” and “leverage”—which tend to appear more often in AI-assisted writing than in traditional academic texts. Comparing these terms against common economics vocabulary helped detect subtle shifts across thousands of papers.

One of the study’s authors explained this as a sort of “linguistic fingerprint” left by AI tools. Because the analysis focuses on language style, it likely underestimates AI use—especially since authors may edit AI-generated text to hide its origins.

Ethical Questions and Policy Challenges

Journal policies on AI use vary widely, ranging from outright bans to mandatory disclosure or no clear guidelines at all. This patchwork raises important ethical issues regarding research integrity and peer review standards.

The study’s authors did not use AI in writing their paper to maintain methodological rigor but note that standard writing aids like spellcheck are common and accepted. The key questions now include:

• At what point does AI assistance become co-authorship?
• Should AI tools be formally acknowledged in research papers?
• How can disclosure policies be standardized?

While the study does not answer these questions, it provides concrete evidence that AI’s influence on scholarly writing is already real and growing.

Toward Responsible Integration of AI in Research

Rather than attempting to ban AI use—which is difficult to enforce—the study suggests establishing clear standards for how AI tools are acknowledged and used ethically. This includes proper citation, transparency about AI involvement, and addressing potential biases inherent in LLM outputs.

One practical benefit is that AI can make research writing more accessible. Non-native English speakers and early-career researchers without extensive editing support might find AI tools helpful to level the playing field. On the flip side, disparities could widen between those with and without access to advanced AI technology.

Beyond Economics: Broader Applications

The language analysis method developed in this study can be adapted to monitor AI’s adoption in other academic disciplines and even in non-academic fields where AI use is widespread. Academic publishing offers a unique, long-term dataset that helps track these changes systematically over time.

For writers interested in gaining practical skills with AI language tools, exploring courses on ChatGPT and prompt engineering can provide valuable hands-on experience.

Reference

The full study can be found in Economics Letters: The adoption of Large Language Models in economics research, 2025.