“Missing Links: A Comparison of Search Censorship in China”, Jeffrey Knockel, Ken Kato, Emile Dirks2023-04-26 (, , )⁠:

[media, FAQ]…While China’s national firewall blocks access to websites, the role that Baidu, Microsoft, and Sogou play in controlling information is in overcoming two of the firewall’s limitations. First, due to the increasingly ubiquitous use of HTTPS encryption, China’s firewall can typically only choose to censor or not censor entire sites as a whole. However, these search engine operators overcome this limitation by selectively censoring sites depending on the type of information that the user is querying. Second, China’s firewall operates opaquely, displaying a connection error of some kind in a user’s web browser. By hiding the very existence of sites containing certain political and religious content, Baidu, Microsoft, and Sogou aid in preventing the user from being informed that they are being subjected to censorship in the first place.

…Most strikingly, we found that, although Baidu—Microsoft’s chief search engine competitor in China—has more censorship rules than Bing, Bing’s political censorship rules were broader and affected more search results than Baidu. This finding runs counter to the intuition that North American companies infringe less on their Chinese users’ human rights than their Chinese company counterparts.

…Moreover, for some social media search platforms, we noticed that, for some queries that did return results, these results seemed to be only from accounts which have received a certain amount of verification or approval. We call this type of censorship in which results are only allowed from authorized sources soft censorship and censorship in which no results are allowed hard censorship (see Table 5 for a breakdown of each platform we discovered performing soft censorship).

To detect this form of soft censorship, for each web search engine, we modified its truism by restricting the results to only be allowed from unauthorized sources. For example, on Baidu, we only allow results from microsoft.com, a site we chose because it is both popular and accessible in China but foreign operated and unlikely to be pre-approved for voicing state propaganda. For Baidu, we surrounded the tested string with site:microsoft.com -(” on the left and “)” on the right in order to transform it into a truism and test it for soft censorship but with the restriction that results were only allowed from an unauthorized source. Thus, for the string “彭帅”, we would test the truism site:microsoft.com -(彭帅), which can be interpreted as searching for any page on microsoft.com not containing “彭帅”. See Table 6 for the rules which we used to create truisms to test each site employing soft censorship.

[The findings suggested that China’s censorship apparatus had become not only more pervasive, but also more subtle. The search engines, including Bing, have created algorithms to “hard censor” searches deemed to be politically sensitive by providing no results or by limiting the results to selected sources, which are usually government agencies or state news organizations that follow the Communist Party’s line.]