An idea for spotting chat-gpt written essays

Suggestion for coping with generative AI: automate the checking of student reference lists against Google Scholar. If the reference doesn’t exist then there’s prima facie grounds for inferring it was hallucinated by chat-gpt. Is there another explanation for this? It’s such a simple identification mechanism that could be done with a basic python script. There would be near zero false positives even if it wouldn’t identify all cases of generated essays. Is this a viable idea or am I missing something obvious? Obviously making it interoperable with university platforms and processes would be more challenging than actually writing the script. But this feels like something which is worth exploring.