Very much not the point in this particular instance. However, if a text were badly in need of proper editing and an AI like Grammarly were used and created the consistent types of errors it usually does produce (notably as "suggestions") a good editor would be able to spot them at once. Conversely, a good editor would do the same job but make actual corrections and Grammarly would scan that editor's work only to make those same "suggestions" as if indicating the editor might have been wrong.
So, in the end, it is not a matter of human vs AI, but a knowledgeable human vs AI. Someone not as knowledgeable might not be able to tell the difference. And THAT is the problem. AI making "suggestions" that a naive or novice writer will see as a "correction."