There were times when it generated “hallucinations”—that is, it would suggest standards that didn’t make sense or align with the lesson. The key to avoiding these pitfalls was to treat AI as part of a human-centered workflow. AI is great for doing the heavy lifting, but I’ve learned that it’s not meant to produce a final deliverable on its own.

Instead of using the basic chat function, I trained a custom GPT (generative pre-trained transformer) for CEFR standards alignment. I started by uploading a PDF of the CEFR standards, including the rubric specifically designed for teaching world languages.

My team had initially developed our curriculum to align with the ACTFL standards, which are widely used in the United States for world language education, but we decided to show how our curriculum also meets the CEFR standards. It felt like a daunting task. That’s where AI turned what could have been a tedious process into something far more manageable—and teachers can do the same thing with their lessons.