Interactive visualizations aid conceptual learning in undergraduate mathematics, yet building them usually requires programming expertise that many instructors lack. This paper adopts a design‑based research approach to propose and evaluate a six‑phase workflow: Foundation, Customization, Mathematical Depth, Application, Accessibility, and Pedagogical Control. Each phase defines a structured prompt that guides a generative AI to produce code compliant with $WCAG\ 2.2$ Level AA standards, and specifies where human verification intervenes.
We investigate whether the structure consistently yields correct and accessible tools, whether it can operate both backward (reverse‑engineering prompts from an existing tool) and forward (generating a tool from a plain‑language idea), and what verification each phase demands. Four deployed tools covering calculus, multivariable calculus, and differential equations are evaluated for mathematical correctness against closed‑form solutions, accessibility through automated and manual screen‑reader testing, and recurring error patterns.
Results show that the workflow generates structurally complete first‑pass tools, but human review of mathematical accuracy, accessibility compliance, and pedagogical fit remains mandatory before proceeding to the next phase. The workflow is platform‑independent and serves both instructors and students.
Blogger's Review: The six‑phase template offers a clear prompting scaffold that turns generative AI into a visualization ally for non‑technical educators, yet human scrutiny is still essential, especially for mathematical rigor and accessibility assurance.