publications

<<previous  |  index  |  next>>

III. Comparative Analysis of Search Engine Censorship

Yahoo!, Google, and Microsoft all argue that Chinese Internet users benefit from their presence, despite these companies' compromise with Chinese government censorship demands. However, this argument would require, among other things, that the services provided by these companies to Chinese users enable greater access to information than they would be able to receive from their domestic Chinese competitors. A comparison by Human Rights Watch of the three companies' search engines with Baidu, China's most popular domestic search engine, indicates that while Google.cn and Microsoft's new "beta" Chinese search engine provide significantly better access to information than Baidu, our experience with Yahoo!'s Chinese search results indicates no better access to information than Baidu.53 We conducted tests on various dates between May and August 2006; results varied, at times even on the same day.

1. Censorship through website de-listing

To illustrate the situation, Human Rights Watch color-coded and tabulated the search results for twenty-five URLs (web addresses) across Google.cn, Yahoo! China at cn.yahoo.com, MSN Chinese “beta” (test) search at search.msn.com.cn, and Baidu, China’s leading domestic search engine (see the chart in Appendix XII). Selection of twenty-five URLs for the URL search comparison chart in Appendix XII focused primarily on politically sensitive websites (such as savetibet.org, Taiwan government, or Falungong), activist sites (Human Rights in China), international news sites (BBC and Time.com), or sites that enable people to share user-generated content or citizens’ media (GlobalVoicesOnline, Technorati, etc.). Websites for a few organizations that the Chinese government views favorably or neutrally were also included (Harvard.edu, Unicef.org, Greenpeace.org) to demonstrate that uncensored results are possible across all services. Here is how the four services break down, ranked according to the number of successful site searches:

  • Google.cn: Seventeen of the twenty-five websites searched yielded the website entered into the search box. Eight of the tested sites were de-listed (showing that the site not only does not appear in a direct search but also will not be drawn upon for any search results in that service). In all eight de-listed cases, a user notice appears in Chinese that says: “According to local laws, regulations, and policies, a portion of the search results do not appear.” Such sites included Radio Free Asia at rfa.org, the International Campaign for Tibet at savetibet.org, BBC News at news.bbc.co.uk, and Human Rights Watch at hrw.org. (See Section IV, Fig. 16.)
  • MSN’s Chinese search engine (beta.search.msn.com.cn): Fifteen of the twenty-five websites searched yielded the actual website entered into the search box. Five were de-listed with MSN’s standard notification to the user: “Some search results have been removed. [click here to] Find out why.” Another five search results did not display the top-level domain originally entered into the search, but did display inner pages from the same website, and in addition provided the same user notification message as in the fully de-listed results. De-listed webpages included such sites as Falungong’s Epochtimes.org and Time.com (the website of TIME magazine, published in the U.S.). (See Section IV, Fig. 11.)
  • Yahoo! China (cn.yahoo.com): Eight of the twenty-five websites tested yielded the actual website entered into the search box. Two yielded sub-domains but not the main domain. One was de-listed (producing no results for the website) with the message "extra results have been filtered." Fourteen were de-listed with a "no results found." (Interestingly, the same search using those fourteen keywords conducted on Yahoo! China from a non-Chinese ISP triggers a browser error. Technically, the reason why this happens is unclear.) On August 8, 2006, a spokesperson for Alibaba confirmed to Human Rights Watch that, since July 27, 2006, at the very bottom of every search results page--regardless of the search subject--Yahoo! China now includes a small line of Chinese text saying: "In accordance with relevant laws and regulations, a portion of search results may not appear." (Also see Section IV, Fig. 5 for an example.)
  • Baidu: Searches conducted on August 9, 2006 inside China produced the same result for every URL or web address search. The browser displays a message saying "you can directly visit [url]." For example, a search for voanews.com will result in the message: "you can directly visit voanews.com." When the user inside China clicks on the link, the user will find that the page is blocked. However it is interesting that in this case, Baidu is letting the Chinese ISP do the blocking, and is not censoring these web pages directly--as the foreign search engines are doing. (Attempts to reproduce this result from outside China consistently trigger browser error. Again, the reason for this is unknown.)

As one can see from examining the URL search comparison chart in Appendix XII, Google.cn and MSN Chinese search yielded substantially more successful results than their Chinese competitor Baidu or Yahoo! China, which appear to provide very similar levels of access. Furthermore, while both Google.cn and MSN Chinese do indeed censor websites, they are transparent with the user in each case that censorship has occurred, even though they fail to inform the user as to why a given URL has been de-listed and under whose authority.

2. Keyword censorship

Human Rights Watch chose twenty-five keywords-twenty politically sensitive terms or names, plus the names of two Chinese celebrities, one company name and two city names-to demonstrate that completely uncensorsed results are possible across all services. These keywords were then plugged into Google.cn, Google.com, Yahoo! China (cn.yahoo.com), Yahoo.com, MSN Chinese "beta" search (search.msn.com.cn), MSN Search (U.S.), and Baidu. The results on Google.com, Yahoo.com, and MSN Search (U.S.) are uncensored for Chinese political terms (though they are censored for copyright violations and child pornography as discussed in Section IV). The four China-based search engines featured two different kinds of censored results: 1) User notification of censorship (user is notified censorship has taken place); 2) Non-transparent censorship (the user is not notified that censorship has taken place).

Google, MSN, and Yahoo! China (as of July 27, 2006) all notify users in different ways that censorship is taking place, although no information is specified as to how many results were blocked, what exactly was blocked, or how they were blocked. The only way to infer answers is to make comparisons with these search engines' U.S.-based counterparts. While Google and MSN give notifications of censorship on results that do indeed include censored results (although in a few cases MSN appears to omit this notification), Yahoo! China's notification is a blanket notice on all pages that censorship is possible in any set of results. Baidu, their Chinese competitor, gave no indication of whether any search results were censored, making it impossible to conclude with confidence that even the most innocuous search hasn't been censored. A more detailed analysis is as follows:

  • Google.cn: Of the twenty-five keywords searched, only three (Microsoft, Dalian, and Paris) did not include the standard censorship notification message at the bottom of the page: "According to local laws, regulations, and policies, a portion of the search results do not appear." (See Section IV, Fig. 15.) For some of the long-standing politically sensitive terms such as “Tiananmen Massacre,” “Li Hongzhi” (Falungong leader) and “Tibet independence,” the majority of results appearing on the first two pages tend to be from PRC sources such as the People’s Daily website, Xinhuanet, Sina.com, and other sites containing articles outlining the Chinese government’s point of view on these issues or people. However, some sites containing perspectives not flattering to the Chinese government or which express support for Li, Tibet independence, or which condemn the Tiananmen massacre do also appear—particularly on blogs or other sites. In other cases, results turned up a great deal of content unfavorable to the Chinese government’s position on the first two pages, despite the fact that the results were still censored, as content appearing at the top of a Google.com search from various well-known human rights and dissident websites does not appear in a Google.cn search.  In the case of “Wu Hao,” the name of a jailed (and recently released) filmmaker and blogger, the first result is the blog written by his sister about his case. Thus, despite results being censored, one can still find a great deal of information via Google.cn search which brings direct exposure to information and ideas that present a different picture of reality than that painted by official Chinese sources.
  • MSN Chinese search (search.msn.com.cn):  Thirteen of the twenty-five searches yielded censored search results in which the user is notified that censorship has taken place (see Section IV, Fig. 10). Eight results appeared to be uncensored. In one case (for “Tiananmen massacre,”) there were no results and no message notifying the user of censorship, just a message saying no results could be found. (See Section IV, Fig. 9.) In three other instances (also highlighted in blue on the keyword search comparison chart in Appendix XIII) it was unclear whether censorship had occurred because there was no notification, but the results appeared in their substance to be more consistent with censored or filtered results in other services. (Note that in the tester's experience some MSN Chinese search results appear to change dramatically depending on what day and time the search is conducted.)
  • Yahoo! China (cn.yahoo.com): All of the twenty-five searches yielded some form of multiple results, and in some cases it is clear by comparing with Yahoo.com that those results are heavily censored. Some searches, particularly "Tiananmen Massacre," and the Chinese term for "Reverse the June 4th verdict" yielded dramatically fewer results than the unfiltered Yahoo.com. A search on "Dongzhou village," where protesting peasants were shot in the summer of 2005, yields only content pertaining to schools, factories and other locations with "dongzhou" in the name, with no content related to the protests and crackdown (see Section IV, Fig. 3). A search on "Wu Hao" (the jailed and recently released filmmaker and blogger) yielded results in the first two pages only about other people with the same name. All results, even on the most innocuous search terms, now carry the notification that censorship may have taken place. (See Section IV, Fig. 5.)
  • Baidu: Of the twenty five search terms tested, Baidu returned zero results for one of those terms, although many others were heavily censored in comparison to other search engine results, without a notice that censorship had taken place.

From the analysis in the keyword search comparison chart in Appendix XIII, one might conclude that, despite their participation in censorship and compromises with the Chinese government, Chinese Internet users have access to significantly more information with Google.cn and the censored MSN operating in China. However, it appears that Yahoo! is censored at approximately the same level as Baidu, the domestic search engine leader.  




[53] The tests below were conducted from the United States on a U.S. Internet Service Provider (ISP), as well as from China on a Chinese ISP, in order to isolate with a high degree of certainty that the censorship discovered as a result of these tests was carried out by the companies themselves, not by the Chinese government or Internet Service Providers at the router level.


<<previous  |  index  |  next>>August 2006