Do ChatGPT, Gemini, Claude, and Other LLM Technologies Make Nations Wealthier? Some Early Evidence

Since the widespread public release of ChatGPT in late 2022, Generative Artificial Intelligence (GenAI) has transitioned from a specialized research area to a significant macroeconomic and social factor. GenAI includes innovations in the field of Large Language Models and transformers.

Their launch has prompted widespread discussion, with governments developing national AI strategies and firms rapidly integrating AI applications. Commentators frequently draw parallels between GenAI’s transformative potential and that of foundational General-Purpose Technologies (GPTs) such as electricity or the Internet, suggesting the potential for revolutionary gains in productivity and economic growth. Conversely, a substantial body of skepticism suggests that AI is subject to overhype, citing difficulties in empirically measuring productivity enhancements and noting the absence of compelling evidence for a macroeconomic breakthrough in aggregate data to date.

Against this complex background, in a new paper, I seek to address the critical and challenging task of empirically quantifying the association between GenAI and income growth. The aim is to offer early evidence-based insights to counter debates frequently dominated by anecdotal evidence, firm-level case studies, and consultancy projections. The central research question of the work is to understand whether countries engaged in developing GenAI experience faster growth in income per capita. Patent data are used to systematically identify and categorize countries actively innovating in both general AI and the more specific domain of GenAI. The underlying hypothesis is straightforward: if GenAI constitutes an economically meaningful technology, countries that are innovation leaders should exhibit a measurable “growth premium” that other countries do not show.

As a first step, the analysis collects data on patent applications at the United States Patent and Trademark Office and classifies them into innovations related to Generative AI (GenAI), general AI (AI), and those that fall in more general digital fields. Overall, these two categories of AI patents together represent one percent of total patent applications in 2022 (see Figure 1). This share is fractional compared to the percentage of digital (ICT) patents that peaked at 15% of total patent applications in the mid-2010s and has fallen since then. Conversely, AI patents are on the rise.

As a second step, using decadal data for a global sample of countries, the analysis seeks to ascertain whether the rate of GDP per capita growth is higher in countries specializing in the development of GenAI technologies, relative to general AI and other digital technologies. Empirically, I worked to identify the effect of GenAI in two ways. First, using the extensive margin effect of innovations, I examined whether countries that engage in GenAI innovation experience faster income growth. Secondly, by looking at the intensive margin of GenAI, I quantified the income growth premium associated with the amount of GenAI developed in the country.

On the extensive margin, I found that countries active in GenAI innovation exhibit a marginally faster growth rate in real GDP per capita. This estimated “growth premium” accumulates to approximately 0.02 percentage points over a decade. On the intensive margin, the estimated contribution of the intensity of GenAI innovation to growth ranges from 0.009 to 0.013 percentage points annually since approximately 2009.

While these numerical estimates are small relative to some of the more optimistic policy and consultancy forecasts, they are not economically trivial at the macroeconomic level, where compounding effects magnify marginal gains over time. Furthermore, the very existence of a detectable positive impact, given the recent emergence of GenAI, suggests that economic relevance is already active. The analysis further indicates that the effect of GenAI often appears stronger than that of “traditional” (general) AI. This suggests that GenAI may represent a qualitative technological step change rather than a mere extension of earlier machine learning trends.

The modest size of the estimated effect should be interpreted not as a failure, but as a reflection of realism dictated by three key factors: (1) GenAI achieved broad visibility only after 2017, with mass adoption occurring even later. The integration and structural changes required for new General-Purpose Technologies to manifest fully in aggregate macroeconomic data typically require years or decades; (2) GenAI innovation is highly concentrated in a small number of economies, limiting the immediate global economy-wide impact; (3) real productivity benefits accrue through widespread diffusion, which requires substantial complementary investments (e.g., skills, data infrastructure, organizational change) that take time and may initially dampen measured productivity.

All this makes this line of inquiry worthy of deep assessment in the coming years.

References

• Venturini, F. (2025) “Generative AI and income growth: Early evidence on global data” Gospodarka Narodowa, Polish Journal of Economics, 3(323), 31–46