Is Austria Top in AI Research? It Depends on How You Count

Who leads globally in AI research? The answer depends largely on how you measure it. A study by Prof. (FH) Dr. Miroslav Despotovic, MA, from HOK demonstrates that different calculation methods can significantly alter international rankings. Looking at the absolute number of scientific publications among the twelve analyzed countries and regions, Austria lands in last place. However, when publications are calculated relative to population size, the country climbs to second place—just behind the Netherlands.

AI rankings are highly sought after in politics and the media to highlight which countries lead in Artificial Intelligence. A primary foundation for these rankings is typically the number of scientific publications on AI topics. Yet, which publications are actually included? How is AI defined? And how are the numbers compared? Despotovic investigated the extent to which such methodological choices impact results by evaluating freely accessible data for eleven countries and the EU from 2015 to 2025.

Austria: Small, Yet in the Top Tier

Counting absolute numbers of AI publications places Austria in last place among the twelve analyzed entities. This comes as little surprise: A nation of around nine million people cannot compete on raw volume with China, the US, or the entire EU.

However, when adjusting publications per capita, the picture changes dramatically. For the year 2024, Austria reaches 97.2 peer-reviewed AI publications per million inhabitants, placing it second—just behind the Netherlands at 98.5. China, while leading in absolute publication counts, ranks significantly lower in per-capita comparisons.

Even when using the number of researchers or research budgets as a reference point, Austria remains in the top tier, ranking between second and fourth depending on the calculation. The Netherlands holds first place across all three relative metrics. It should be noted that these comparisons are limited to twelve selected countries and regions and do not represent a complete global ranking.

Four Decisions That Reshape the Results

Before a number appears in a ranking, decision-makers must define what is being counted. The study examines four key methodological decisions. Despotovic altered each factor individually to compare its impact on national rankings:

1. What counts as a publication? Only peer-reviewed journal articles, or also conference papers and preprints published on the internet? Conference papers play a crucial role in AI research. Depending on which publication types are included, different countries take the top spot. For example, in 2020, three different leaders emerged depending on publication type: the US, the EU, or China.
2. What qualifies as AI? Depending on whether the topic is defined broadly or narrowly, the number of captured publications can double. However, the overall order of countries remains largely consistent.
3. Which data source is used? The choice of database also influences outcomes. Despotovic compared OpenAlex, an open literature database, with the Emerging Technology Observatory (ETO) at Georgetown University. ETO reports roughly 1.4 times more AI publications for Austria than OpenAlex, and about 3.1 times more for South Korea. A likely reason lies in differing indexing methods: ETO includes preprints and employs its own AI classifier, which captures subfields like computer vision, natural language processing, and robotics more broadly. Countries with strong research focuses in these areas, particularly in Asia, appear more prominent as a result. However, these data sources are not entirely independent, as OpenAlex feeds into ETO’s database.
4. Absolute or relative metrics? The most significant shifts occur when comparing absolute publication counts against population, researcher count, or research budgets. This is precisely where Austria’s standing changes most drastically: from twelfth place in absolute numbers to second place in publications per million inhabitants.

What Does This Mean for Austria?

Anyone discussing Austria’s strength in AI research should always clarify which metric is being referenced. In terms of absolute publication output, a smaller nation naturally trails large research nations. Yet relative to population size or research resources, Austria belongs to the leading group among the analyzed countries and regions. Together with the Netherlands, Austria demonstrates that smaller European nations deliver remarkably productive contributions to AI research relative to their size.

For businesses in Austria, this is encouraging news: Domestic AI research is highly productive relative to the country’s size and research funding. This provides a strong foundation for expanding collaboration between industry and research institutions to turn scientific insights into commercial innovations.

At the same time, the findings highlight the limitations of international research rankings. Publication counts indicate how much scientific research is being published, but say little about its quality or real-world impact. Every publication is initially counted equally, regardless of the journal or conference in which it appears or the citations it generates. Factors such as computing capacity, infrastructure investments, skilled talent, and commercial implementation were not covered in the study.

The study therefore recommends full transparency regarding underlying calculation methodologies in international AI rankings. This includes disclosing the data sources used, the definition of AI, the publication types included, and whether absolute or relative metrics are being compared.