AI and education in African national AI strategies: An analysis of 18 countries

National artificial intelligence strategies and policies assign education responsibilities both for developing AI capabilities (education for AI) and for adopting AI in educational services (AI for education). This study examines how these responsibilities are articulated and supported in a bounded corpus of 18 African national policy documents. A descriptive qualitative document analysis consolidated 275 evidence records, retaining source passages, locations, working translations, statement types and uncertainty notes. Document-level coding distinguished learning about AI, identifiable educational AI applications and educator preparation; a framework-based thematic synthesis examined delivery arrangements, inclusion and safeguards. Structured AI learning was identified in all 18 documents, named educational AI functions in 12 and educator preparation in 14. These counts include proposed and illustrative measures and do not indicate implementation; restricting them to proposed actions gives 17, 9 and 13. Educational applications ranged from personalised learning and tutoring to student profiling, admissions and credential verification. Supporting arrangements included teacher development, laboratories, scholarships, partnerships, institutional responsibilities and monitoring targets, but these elements were not interchangeable evidence of a complete delivery plan. Counter-examples to any general claim of weak specification included a costed education programme in Mozambique’s draft and proposed algorithmic bias audits covering education in Côte d’Ivoire. The corpus assigns education a dual role in AI development while specifying the connections between reforms, resources and protections unevenly. The contribution is an evidence-linked account of policy design, not a ranking of countries or an assessment of educational outcomes.