Sub-Saharan Africa contributed just 0.83% of total global AI publications between 2013 and 2024, according to the AI Index 2026 Report published by Stanford University’s Institute for Human-Centred Artificial Intelligence, highlighting a research gap that analysts say undermines the region’s ability to develop homegrown large language models and participate meaningfully in global AI innovation.
The figure stands in stark contrast to North Africa and the Middle East combined, which contributed 4.27% of publications over the same period. East Asia and the Pacific led all regions with 26.42% of AI publications in 2024, with China alone accounting for 17.8% of that total. European and American publications accounted for 19.5% and 12.6% of global citations respectively, while Chinese AI research represented 20.6% of all citations in 2024.
The report’s authors note that a relatively small number of countries account for a disproportionate share of global AI research activity, a pattern that reinforces existing inequalities in who shapes the technology’s development.
The finding carries direct policy implications for African governments. As South Africa, Kenya, Nigeria and other nations draft national AI strategies, analysts say policymakers must address the continent’s research deficit alongside governance frameworks and infrastructure investment if they want AI to serve local needs.
Overall, AI publications more than doubled over the decade studied, rising from roughly 102,000 to approximately 258,000. AI research now accounts for 40.9% of all computer science publications tracked in OpenAlex. Journals claimed the largest share of publications at 62.8% in 2024, compared to 23.8% for conferences. The number of AI doctoral graduates in North America increased 22% over the past two years, with many new graduates choosing to remain in academia rather than move to industry despite high salaries in the private sector.
On the model performance front, the United States continues to lead in top-tier AI models. The report notes that as of March 2026, Anthropic’s Claude holds a 2.7% lead over competing models. China has narrowed the gap significantly, with DeepSeek approaching U.S. model performance levels in early 2025. “The U.S. still produces more top-tier AI models and higher-impact patents, while China leads in publication volume, citations, patent output, and industrial robot installations,” the Stanford report said.
