This is a description of the work: https://citizenlab.ca/2022/05/bada-bing-bada-boom-microsoft-...
The work is quite interesting to read (having been on the other side of systems like this) but it uses statistical techniques like fisher tests in a fairly naive way. My next complaint is that they use the term censorship. I prefer to use the term "suppression of legitimately interesting results to further a political or business need".
At some point the paper devolves into speculating abotu the complex serving systems that produce suggestions at runtime (they are a function of thousands of feature variables). Probably the best readers of this paper would the (US-based ) technical lead and product manager for Bing, who should read it in detail and plug all the obvious bugs that were found.