Global AI Chatbot Usage Varies by Country Income and Language
A new study analyzing anonymized interactions with a widely available free AI chatbot reveals significant differences in how early adopters use generative AI across countries. Schooling is the most common domain of use, especially in low-income countries, where a strong inverse association exists between schooling-related use and GDP. Leisure-related use, conversely, is positively associated with higher country income. Language also shapes usage: English-language interactions are overrepresented in regions where dominant languages were poorly served by existing models during the study period. The research suggests that improving multilingual performance could determine whether AI widens digital divides or enables leapfrogging.
Key facts
- Study uses large-scale dataset of anonymized, de-identified, privacy-scrubbed interactions with a free AI chatbot.
- Schooling is the most common domain of use in most countries, particularly low-income countries.
- Strong inverse association between schooling-related use and country-level GDP.
- Leisure-related use is positively associated with country-level income.
- English-language interactions are overrepresented in places where predominant languages were not well-served by existing models.
- Improving performance across languages may be key to whether AI expands digital divides or enables leapfrogging.
- Research empirically characterizes differences in early adopters' usage across countries.
- Study period covers time when some languages were poorly served by AI models.
Entities
—