AI Chatbots – World Bank Blogs
These days, one of the countless tabs always open on my internet browser is ChatGPT. Just in the past week, I have used it for Writing (“edit this email to my daughter’s daycare…”), Seeking Information (“which footmuff is best for our stroller”), and Practical Guidance (“suggest solids for sending to daycare”). While not all 700 million users of ChatGPT are new parents like me, according to a recent NBER working paper, my usage patterns are quite typical: almost 80% of all consumer ChatGPT messages fall into these three broad categories, and non-work messages represent more than 70% of all messages.
ChatGPT (like Claude or Gemini or Copilot) is an AI chatbot that has been trained to generate text/ images/ speech – by a large language model (LLM) running in the background – in response to a user prompt. Within 32 months of launch, ChatGPT has reached roughly 10 percent of the world’s adult population, a diffusion rate faster than any prior consumer technology!
Who are the users?
Chatterji et al (2025) analyze global data and identify usage patterns by age, gender, country of origin, education, and occupation between November 2022 (when ChatGPT was launched) and July 2025.
Implications
Access to these gen-AI chatbots is likely to have fundamental implications for how we work and learn in the future. For instance, 10% of all ChatGPT messages were requests for tutoring or teaching. Moreover, Writing accounted for 40% of work-related messages in June 2025.
Does this mean that in the long run students and workers will rely on AI for writing and problem-solving instead of developing their own abilities? Will this weaken investment in communication and critical thinking—soft skills that are highly valued on the labor market because they are central to leadership, collaboration, and innovation, but that are hard to automate? Perhaps education systems will need to adapt by teaching students skills that complement, not compete with, AI tools? If so, what are these skills? It may also get harder to assess genuine learning if AI is doing much of the work behind assignments and exams, which could reduce the signaling value of degrees in the labor market and lower incentives to complete formal education. Over time, this dynamic could lead to a mismatch between the skills people acquire and those that the economy truly needs, with individuals focusing on areas easily automated by machines rather than building deeper capabilities.
These are open questions and it remains to be seen how gen-AI will influence education and skills development at scale and in the long run. That said, governments and policymakers need to act now to ensure that they are ready for this new technology. How should we think about the potential of AI for economic development in LMICs? This is something that the next World Development Report will explore. Stay tuned!
Thank you for choosing to be part of the Development Impact community!
Your subscription is now active. The latest blog posts and blog-related announcements will be delivered directly to your email inbox. You may unsubscribe at any time.
Senior Economist, Development Research Group
