how many of them spend more time revising what AI writes to mask detection than they would've actually writing their papers

I am so frustrated by this. The one assessment I had to mark had a required reading of about 1-2 pages. It would have taken students 10 minutes to read the actual paper. But most if not all of them relied on revising AI generated papers and other students' papers (yes, even when AI is there, students still copy from other students). So, they spend MORE time revising AI slop and really bad answers, rather than just read the actual work. I am not even sure how this logic works.

built into a browser or a plug-in that analyzes everything we read

This will be such a good help. But in terms of Universities, they still need to bite the bullet and just implement this. The institutions I work for all rely on Turnitin, and each and everyone of these institutes tell us "these software are not reliable". So, in my opinion, they just need to bite the bullet and tell us which ones to use, so that we can actually learn students also how to use AI in a productive and responsible manner.

It feels like we markers and lecturers are playing more the role of "police" than actual teaching and assessing.