Kenya ranked first globally for ChatGPT usage in a survey published by DataReportal in July 2025, according to TechCabal. The report found that 42.1% of Kenyan internet users aged 16 and over had used the tool in the previous month. South Africa followed with 34.3% and Nigeria with 27.9%. This ranking presents a stark contrast to Microsoft’s data for the first quarter of 2026, which placed South Africa at the top with 23.1% of its working-age population using generative AI tools. In that same period, Nigeria recorded 10.1% and Kenya 8.7%.
The reversal in national rankings is driven by the different populations each survey measures. DataReportal’s figures are derived from GlobalWebIndex surveys that target only internet users. In contrast, Microsoft estimates usage across the entire population aged 15 to 64, regardless of internet access. TechCabal notes that this distinction is critical because internet penetration varies significantly across the region. DataReportal’s Digital 2027 reports indicate that internet access reaches 79.5% of South Africa’s population, compared to 40.6% in Kenya and 47.5% in Nigeria. Consequently, a high percentage of internet users in a country with low overall connectivity may result in fewer total users than a lower percentage in a highly connected nation.
Methodological differences extend beyond the denominator. Microsoft’s data is primarily sourced from Windows desktop and tablet platforms, requiring assumptions to account for mobile device usage. TechCabal reports that Microsoft’s technical paper acknowledges these assumptions are difficult to validate. This is particularly relevant in Africa, where mobile devices are often the primary gateway to generative AI. Furthermore, Microsoft pools data for some markets; Nigeria’s 10.4% second-quarter estimate is shared with Ghana, Benin, and Togo, making precise national comparisons challenging.
These statistical divergences have practical implications for businesses and governments. For startups, high ChatGPT familiarity among internet users may indicate a viable customer base for AI writing tools. However, TechCabal argues that this metric does not reveal whether users find the tools valuable enough to pay for them. Governments face similar complexities. A ministry deploying an AI assistant must consider not just familiarity with chatbots, but also digital literacy and device accessibility. The distinction between a student using AI for a single assignment and a trader using it weekly for business operations remains obscured by broad adoption statistics.
