@mguhlin In Bilingualism classes in Japan, I can require students to show what they have learned from this class, which is based on reliable research, and to discuss their own experiences in summary, reflection, and response papers. In such ways I can discern how much of students' writing is their own thinking and analyzing. Findings from the discipline of bilingualism are dwarfed by the amount of common misconceptions about bilingualism, which would presumably be reflected in AI databases, from blogs and such sources in Japanese, where students find mostly stereotypical, anecdotal, or prejudiced information for papers, despite my warnings to use reliable sources. I am wondering how applicable this approach could be for other subject matter areas.
Among the limitations are that the English is too much and too difficult for L2 learners in most academic sources, such as journals. When students use automatic translation, they have probably thought about the content in Japanese, but they can paste the auto-translated English into papers without fulfilling the language-driven mandate to balance the content-driven approach. I do not believe in content-based EFL as a pretext to mask a language-driven agenda, since the students are enriched by the transferable knowledge they gain in L1 as well as in L2.
Writing reflects thinking, so I discourage writing and speaking that relies on machine translation or AI. The result might be obvious, but we cannot accuse and police students.