Your 9% means what? I could not find any differences between chinese/us google in results from "abreast".. do you have ANY example from your common words list which actually gives different results?
Not on 10's page, I mean, but on the first or second one?

Are you saying the results are skewed with Chinese government flavored results or that these words are being monitored by the Chinese government. Kinda makes for a funny image a bunch of guys in Red Army uniforms sitting around staring at the populace's search results.

Interesting ... I wonder what the percentage would be if you tried that in a few other google.country-codes?

I.e. isn't there some censoring done in Germany and France?
alek

P.S. Assuming I understand this correctly, might it be slightly better worded if you said that 9% of the searches returned web sites in the top-10 results that Google has agreed to censor in China?

I.e. it's not the word itself that is tripping some censorship filter – it's a website out there that happens to be returned in the SERP's – right?

By the way. does anyone know of proxy servers *in China* that sell accounts? That is, I'm not looking for the list of proxy servers which just happen to be open, there's lots of them. Rather, I mean *Chinese* proxy servers which have it as their business to let people use them.

Answered my own question

http://proxify.cn/

Anonymous 1:
> Your 9% means what? I could not find any
> differences between chinese/us google in
> results from "abreast".. do you have ANY
> example from your common words list
> which actually gives different results?

Please read the footnote. I simply checked for the appearance of Google's disclaimer on result pages which admits that the result is censored. (“According to local laws and policies some search results are not showing”.)

Anonymous 2:
> Kinda makes for a funny image a bunch of
> guys in Red Army uniforms sitting around
> staring at the populace's search results.

Yes, internet cafes are monitored. So are blogs. Sometimes, email accounts, at least from what we heard (who knows for sure?). Result sites are blocked by the Chinese gov't, if not already by the search engine. A lot of things are monitored by the many, many Chinese internet police men. It's a rather sad image. The monitoring of the censorship blacklist in this example is implemented by Google, of course, based on gov't request.

See this Wikipedia article for more information:
http://en.wikipedia.org/wiki/Internet_censorship_in_mainland_China

Or read this Amnesty International press release:
<<Amnesty International records the cases of at least 33 people who have been detained or imprisoned for offences related to their use of the Internet. They range from political activists and writers to members of unofficial organizations, including the Falun Gong spiritual movement.

One of the longest sentences has been passed against a former police officer, Li Dawei, (...), who has been sentenced for 11 years in prison for downloading articles from Chinese democracy websites abroad. All his appeals have been turned down.>>
http://web.amnesty.org/library/Index/engASA170562002?OpenDocument&of=COUNTRIES%5CCHINA

Alek:
> P.S. Assuming I understand this correctly, might it be
> slightly better worded if you said that 9% of the searches
> returned web sites in the top-10 results that Google has
> agreed to censor in China?
>
> I.e. it's not the word itself that is tripping some censorship
> filter – it's a website out there that happens to be
> returned in the SERP's – right?

Exactly, and I tried to explain that and all other important points in the long footnote. "901 – 9% – of the words checked returned censored results, with 1 or more sites missing from the top 10 results."
"Also note that Google censors sites, not words"
There are several ways to determine the "amount" of censorship:

- do all possible (endless) search queries and then extract a blacklist of domains by comparing Google.cn and Google.com – to do this, it's not enough to check for sites missing on Google.cn, because different countries often have different results. Instead, you need to look at both the disclaimer on Google.cn, and then query Google.cn for all domains in Google.com using the "site" operator (with potential marginal errors as the Google.com top 10 may not include any of the favored but censored Google.cn country-specific results).

- ask Google or the Chinese gov't for a blacklist of censored domains. I just emailed Google Press asking for this, let's see what they reply.

- ask Google for a list of all searches performed over a given week by Chinese users and determine the ratio of censored search results; as I mentioned in the footnote, this ratio might be heavily skewed because a) Chinese people are being watched by an internet police so they have reason to be afraid to enter politically "sensitive" terms in the first place, thus not making those appear on this list, and b) Chinese people may not enter certain "sensitive" terms having learned that the result page will either stall on them or return censored results anyway.

- probe with a sample set of English words, Chinese words, popular phraes (that I've done in a previous post link from the footnote), probe for words Google says they want to work (e.g. "bird flu", which Peter Norvig says works censorship-free but it doesn't) etc.

Does that make probing with the last approach, the word list, less relevant? No, I think it's one of the many imperfect but relevant approaches we can use to probe the scope of Google.cn censorship. We also need to keep in mind that while Google may censor on blacklists only, it is very likely that the Chinese gov't derived certain blacklists by searching on search engines. In other words, when they enter say "bird flu" in Google and see xyz.com, they might simply send the blacklist addition of "xyz.com" to search engines. In this way, they're indirectly censoring *search queires* specifically. I leave that question open for the current post's probe.

On another side-note, while I *think* it's not very likely that people perform single-word queries only (I also mentioned that in the footnote), those who know about SEO know that if a site ranks well for e.g. "communism", it will also rank well for search queries in the neighborhood of that word, e.g. (just as an example!), "communism criticism" or "communism discussion" etc.

Philipp, I understand what you did, but I also don't see a difference in the results for "abreast". I see the censorship disclaimer on the Chinese page.
But I don't see what result is missing when compared to a similar page without the censorship disclaimer.

Maybe the censorship disclaimer is being triggered for some wrong reason?

> Maybe the censorship disclaimer is being
> triggered for some wrong reason?

I don't think so. As we all know, different countries often show different results in Google. I can't even compare the specific results for "abreast" because it's one of the searches triggering a "cluster" result in Google.com, that is, half the results are actually for "dictionary" not "abreast". Or take a search for "Chinese", another self-censored Google result judging by the disclaimer; the Google.cn search result includes a lot of Chinese web pages not appearing in Google.com, simply because (as we all assume) a country-specific search result is always pushing country-specific or local language sites up higher in the rankings. So all we know is that "from the top 10 pages Google.cn *without* censorship *would've* displayed, 1 or more pages are missing on the *actual* Google.cn with censorship". We might be able to find out just which sites are missing by checking the Google top 100, but even then, we might never find the blacklisted domain because Google.com might not rank it high.

Or to put it differently, more formalized. I'll do it for a "top 3" to simplify writing.

----------------------
Google.com ranking for "X":
1. a
2. b
3. c

Google.cn "uncensored pre-ranking" for "X" (whatever Google determins to be top, but doesn't yet output to the user):
1. d
2. e
3. f

Google.cn censored ranking for "X":
1. d
2. f
3. g

As you can see, "e" is missing on Google China, yet we don't have any way to determine what "e" is by looking at the Google.com results for the same query.
----------------------

Or, another example, this one where we do have a chance to reverse-engineer the missing site to some degree:
----------------------
Google.com ranking for "X":
1. a
2. b
3. c

Google.cn "uncensored pre-ranking" for "X":
1. d
2. a
3. b

Google.cn censored ranking for "X":
1. d
2. b
3. c

As we can see, "a" wouldn't have been #1 even in an uncensored Google.cn, because "d", a more country-specific optimized site, would be on top. *However*, "a" would be the #2 spot even on Google.cn – yet in the actualy censored output, "a" is completely omitted, and "b" and "c" move higher up. Using a "site" operator search, we can pinpoint "a".

PS: As an example for the latter type of search censorship, search for e.g. "essay" (a word picked randomly from the post's list, we can use many other words) and compare Google.cn with Google.com results. You'll notice that while the results are different in many "harmless" ways, one of the sites in the Google.com top 10 – a site hosted on geocities.com – is not only missing from the Chinese result, but also returning a Google.cn censorship disclaimer when searched for with the site operator. In other words, there are approximately 9 million pages missing (we don't know for sure, it's one way to derive this number) on Geocities alone, and of course pages hosted on Geocities rank in the top 10 for a variety of queries.

Google.cn [essay]
http://www.google.cn/search?hl=zh-CN&q=essay&btnG=%E6%90%9C%E7%B4%A2&meta=
Top 10 results show with censorship disclaimer

Google.com [essay]
http://www.google.com/search?hl=en&q=essay&btnG=Google+Search
Top 10 results show without censorship disclaimer, including several sites missing on Google.cn but at least one that also misses on a Google.cn site search

Google.cn [site:geocities.com]
http://www.google.cn/search?hl=zh-CN&q=site%3Ageocities.com&btnG=%E6%90%9C%E7%B4%A2&meta=
0 results, censorship disclaimer

We don't know for sure why Geocities had been blacklisted by the Chinese gov't, but we can derive several theories, including that there's a lot of grassroot, mirrored "controversial" stuff on Geocities. E.g. a search for [site:geocities.com mao terror] returns ~10,400 results.

This is getting overly complicated. Here's the summary:

* Let's say I am against China and write bad things about China in my blog.
* My blog is blacklisted by the chinese.
* I then write about "cute little ponies",
* My blog entry gets indexed
* Phil searches on google.cn for "cute little ponies"
* My blog is one of the results, but has to be removed
* Google prints out disclamer
* SURPRISE: Google is censoring "cute little poines"

.....

This means nothing to me, it just means that people blocked by china have written about the 9000 words mentioned.

Exactly, the word list is not necessarily containing "sensitive" terms only. No one ever suggested that either.

(It's 901 words, not 9000 by the way.)

Off by 01 ... common error among computer scientists...

To take your "pony" example, note that to appear censored in the top 10 you'd have to first make it in the top 10 for that word. You can't simply write about "pony" and then search for the word and see yourself censored (if your domain would be blacklisted), you need to be the one of the top 10 sites of the millions of sites available on the subject that Google determines "most relevant" before it then applies the censorship. ("Most relevant" includes backlinks containing that link text, for one thing.) Remember, Google won't tell you if "sites containing that word have been censored", because then the list presented in the post would consist of 99-100% of the words, not 9% (the 25 million censored pages on news.bbc.co.uk *alone* might cover the dictionary of 10,000 words – I could test this of course). They'll only tell you "1 of the results of the 10 most relevant for this query is censored".

So, your example is a bit misleading in that sense, but your conclusion that indeed a lot of "random" words are censored (not exclusively random words, but a lot) is correct. And yes, we can also conclude that the site blacklist is so imprecise (it includes domains, not page URLs!) that there's a lot of "irrelevant" censorship going on... or let's say, irrelevant to the Chinese gov't, not necessarily the blogger/ BBC journalist/ Geocities site owner who got kicked out of the system. So yes, if you have a pony site, better not write any single paragraph against the Chinese gov't, as it might take the rest of your site down (unless you favor truth over rankings, that is).

Philipp, right, I understand there's flux in the results, differences, etc. I was checking different languages too, in order to turn off the annoying dictionary results in the middle.

What confused me was that for "abreast", the censored and uncensored results on page 1 had the same items.

It looks like what happened is that for "abreast", an item which *would* have been around #9 or #10 on the censored results ended up as #11 or #12 on the uncensored results.

It turns out to be a geocities site, this one, [verifiable by searching ["two abreast"]

Côte-à-côte/Two Abreast
l'équipe montréalaise des survivantes ducancer du sein.
http://www.geocities.com/colosseum/park/9839/Ehomepage.html

I have trouble getting worked up about what Google is doing with respect to search results for China.

It's China. Let's either invade the place to establish democracy or simply stop whining about how they run their country already.

Phillip a small question, who decides what to censor and what not to.
I know Google "Self-Censors", so does that mean that Google blocks out any thing IT thinks is unappropriate or does the Chinese govt. give Google a list.

Good update Seth!

Sohil, from testing we can see that Google censors domains and sub-domains.
For example, [site:bbc.co.uk] returns many results in Google China whereas [site:news.bbc.co.uk] returns 0 + censorship disclaimer. (Presumably that's because BBC does have a self-censored Chinese version, so only their international news are censored.)
At other times, the domain is censored, e.g. geocities.com.

Assuming that Google isn't Darth-Vader-evil – as in: hey, we hate Human Rights and freedom of speech too, let our engineers figure out what to censor – what conclusions can we draw? I think the most obvious is this: Google gets handed over a list from the Chinese gov't which show pages that must never appear on Google.cn. Google then either simply accepts the list or discusses this or that domain on the list*. They then calculate "normal, but localized" Google.cn results. They then compare if any domain on the results is on the blacklist, and if so, remove the result (and move a lower result higher up) and display the censorship disclosure.
Technically, as the blacklist may be of "use" to other search engines, there may be automated ways of transmitting the list from a Chinese gov't server to Google – Yahoo – MSN etc.

---------
*During the Google press day, I asked whether Google ever discusses items on the blacklist, but the answer given was evasive. I would like to know if Google goes like "Can you please check again if Human Rights Watch *really* needs to be censored? Can we find a compromise here?"
Because there's a (moral) catch-22: if Google does this for only a single site ever, then they are personally responsible for all the other censorshipped sites they let through (because they chose what sites deserve their arguing for them). And if they never do this for any site on the blacklist, then they are responsible for not even *trying* to keep the censorship down on their Chinese search engine. In either way, we can see Google does get involved in some way here, and involvement means responsibility.

I wonder if there is something like a large "list of important domains" (geocities.com, cnn.com, etc., with thousands of entries)? With that list, I could check which of the domains are blacklisted in Google similar to this post.
http://blogoscoped.com/archive/2005-01-15-n50.html

Open Directory Project

http://dmoz.org/

Oh, they don't like angelfire.com either (another free webhosting pages site)

I'm including an audio version of this post:
http://blogoscoped.com/enclosure/censored-2006.mp3

What TTS soft did you use?

I bought TextAloud yesterday. They have a trial version but the voice I used is "AT&T Crystal16" which was only included in the paid version.

... disappoint, disappointing, disappointment ...

I like the Audio version..

Your MP3 reminds me of Radiohead's Fitter Happier

"more productive
comfortable
not drinking too much
regular exercise at the gym (3 days a week)
getting on better with your associate employee contemporaries
at ease
eating well (no more microwave dinners and saturated fats)
a patient better driver
a safer car (baby smiling in back seat)
sleeping well (no bad dreams)
no paranoia ..."

I wonder about the details of Google's role is in this censoring business. We have this sad image of a team of Chinese censorship police sitting around monitoring search queries, but does Google employ people to work directly with them? Who is the programmer who set up the means within the Google.cn search system to allow them to censor words as they see fit?

It'd be interesting to see an interview there...maybe one day when they're an ex-Google employee.

That audio was pretty decent.

Flipping around, a lot of the differences seem to be due to the BCC and geocities.

They don't like this one either:

http://www.wenxuecity.com/

"wenxuecity.com is the largest overseas Chinese community portal in the world. For the Chinese community around the world, it covers the most important issues and significant trends affecting their lives. Wenxuecity.com's mission"

Nice text2speech, it confirms once more the vision I wrote down a couple of hours ago about the future of our dear Net: http://blogoscoped.com/forum/40304.html#id40400.

Ah, I knew someone had already done a list, ONI has some URLs:

http://www.opennetinitiative.net/studies/china/cn_long.html

I've downloaded the Dmoz.org data file now and will see how it can be used to further probe Google.cn's behavior.

I can confirm Wenxuecity.com's importance at least for my Chinese friend. She used to read it everyday but turned away slowly as she suspected "pro gov't" behavior.

As founder of Project Censored, which has been exploring news media censorship for more than 30 years now, I find it odd that "censor" or any form of the word is not included among the 901 words listed.

I checked the term Carl. "Censor", "censored", "censorship" and "censoring" are indeed uncensored on Google.cn. The first censored result for "censorship" is in Google results 41-50 – the first censored result for "censored" is in results 11-20. But here are some searches in the neighborhood of censorship resulting in 1 more or missing sites in the top 10 (according to Google's disclaimer):

   chinese censorship
   chinese firewall
   great firewall
   chinese internet police
   jailed blogger
   jailed dissident
   china censors
   human rights censored
   human rights online
   journalist jailed
   geocities censored
   mao censorship
   propaganda department
   chinese news
   chinese newspaper
   website blocked
   anti-censorship proxies
   google censored
   falun gong censored
   bird flu coverage
   tiananmen censored
   etc. ...

Sample search:
http://www.google.cn/search?hl=zh-CN&q=chinese+censorship&btnG=%E6%90%9C%E7%B4%A2&meta=

Irony – the "censorship" item is

http://news.bbc.co.uk/2/hi/technology/4973114.stm

they always say "Great FireWall", said it blocked google.com,and things related it. now my gmail seemd said to me "you disconnected from me", but google.cn shows he still stand with me.we cant visit sites like "*.blogspot.com",and things related wiki the same. and when i use some web proxies thier hosts outside of china to go to "google.com", it still redirected to "google.cn".

Google.cn may be beated down by baidu.com.
In china all the people used to using baiud.com not google.cn.
I think google is not special on chinese words searching.
hoopz.biz

Google, Yahoo and MSN censor in America. Googel wouldn't run ads for my
websites, www.ncjusticefruad.com and www.chinaisevil.com. The reasosn they
gave for not running the ads were all lies! MSN and Yahho wouldn't run as for my china site. I filed suit aginst them (Langdon v. Google, 06-cv-319, US DIst. Crt., Del.). You can get a copy form the wired blog, June 8th, 2006, or by searching "Christopher Langdon v. Google" on Yahoo or mSN. Many of the results have links to the Wired blog. You can get a description of Google's censorship in the US at www.chinaisevil on the "Googlegag" page.

   Sincerely, Chris Langdon, qiologyaol.com

Is en.Wikipedia.org censored for you Shelleh? (Please don't answer if this could get you in trouble.) I'm asking because Wikipedia is one of the constantly good, as of yet uncensored sites on Google.cn when researching human rights topics.

Edit: I saw this post being slightly misunderstood in several places so I'm changing the title to clarify the distinction between a word being censored, and the search result for a word being censored. I am changing the title from "Google Censorship Word List" to the more descriptive "Words Returning Censored Google Results". (In the original post, I made this distinction clear in the introductory paragraph, the footnote and the comments, but I think it wasn't too clear in the original title.)

Good edit – I didn't want to make an issue of it, but I can see the reason for the misunderstanding.

So, I asked Google for a list of domains and got this form response:

"Can you tell me the list of self-censored domains in Google.cn?"

<<Thank you for your note. It is Google's policy not to censor search results. However, in response to local laws, regulations, or policies, we may do so. When we remove search results for these reasons, we display a notice on our search results pages. Please note: For some older removals (before March 2005), we may not show a notice at this time. We're unable to provide a complete list of these sites at this time.

Regards,
The Google Team>>

Yeah – the thing there could be one page on a blocked site that contains every one of those words (and theoretiocally it could be the only page Google censors) – I know it isn't though.

to PPhilipp:

no trouble. i trust truth and my country .

it not censored, it be blocked. i cant visit zh.wiki nor en.wiki or likes, unless use proxy.

Shelleh can you visit this link?
http://en.wikipedia.org/wiki/Main_Page

I hope google will retire from the chinese market. "Dont be evil" was written in the google "mission" at the start. now they can write, "be evil.. only in china for now... then in the rest of the world"

2006torino.blogspot.com

Hi everybody,

I try look at google cina, I try with www.google.cn and with www.google.com.cn and it always turns www.google.it from wich I am

What to do?

have nice time

Livio

Interesting if not scary. Scary because very interesting.
I wish I knew what happens in North Korea as well...

> What to do?

See here:
http://blogoscoped.com/forum/40886.html#id40914

i see someone got a good idea, to write a essay with all of the words that apeared in the "list", here is the url: http://underh2o.org/blog/2006/06/20/%e6%95%8f%e6%84%9f%e5%ba%a69%ef%bc%85/

you guys are just RIDICOLOUS!

i live and work in China :

1) first of all, Google (here called "Gu Ge") has only 5% of the search engine market here, any other use Baidu.com, Yahoo.cn, and other smaller chinese SE.

2) secondly, if you don't like the filtered results of google.cn you can just use google.com, .co.uk, .de, .fr, and so on, so to get the unfiltered results.

3) third, it seems you missed the fact that the average chinese do NOT speak english therefore he'll hardly give a shit if BBC or geocities dont work.
it's not that as you wrote "they don't know to have firewalls" but that they plainly don't give a shit.

4) blogs : blogspot banned at DNS level, but 1000s of other free blog providers works fine and fast, so what ?
the government only care about CHINESE blogs, not about english language stuff.

5) apart the IPs blocked by DNS, anything else works fine even using the lamest CGI proxy like anonymouse.de ... TOR works fine too.

6) BBC is the only media censored, since CNN, IHT, ABC, Lemonde, and any other famous newspaper work fine.

7) yeah, DNS and filtering changes from District to District, and from Region to Region so we can't talk of a global unified filtering system.

8) so woah there's also some of you wanting to write an essay on the crap you wrote.. you guys must be brainwashed.
come here and see the reality, kids.

> 1) first of all, Google (here called "Gu Ge") has only 5%
> of the search engine market here, any other use
> Baidu.com, Yahoo.cn, and other smaller chinese SE.

... which are also self-censored (and maybe use similar blacklists, who knows).

> 2) secondly, if you don't like the filtered results of
> google.cn you can just use google.com, .co.uk,
>.de, .fr, and so on, so to get the unfiltered results.

Yes, there are many ways to circumvent censorship. Some less, some more dangerous. But we also heard of cases where Google.com was completely blocked, while .cn was still available. Even if that's not the case, this blog covers Google – when Sergey Brin starts handing little kids on the street poisoned chewing gums, I will not ignore this simply because there are ways from keeping your kids off that particular street. (Ignore the details of that analogy :))

> 3) third, it seems you missed the fact that the average
> chinese do NOT speak english therefore he'll hardly
> give a shit if BBC or geocities dont work.

And it seems you missed to read the footnote:

<<The resulting list of words is neither a complete list of censored search queries search queries resulting in censorship – which is pretty much endless, as words can be combined, and written in different languages – nor is the list in any relation to how common or uncommon a search is, in particular as it’s not a Chinese word list, and in particular as people don’t just enter single-word queries. (But even when we’d have Google’s statistics of the most common searches on Google.cn/ Google.com from China, that list wouldn’t be too meaningful – it won’t list the kind of queries people are too afraid to enter, or don’t enter because they assume censored results anyway, which as we can imagine would be a word list with a higher overlap with censored results.)>>

> 4) blogs : blogspot banned at DNS level, but 1000s
> of other free blog providers works fine and fast, so
> what ?

See #2.

> 5) apart the IPs blocked by DNS, anything else
> works fine even using the lamest CGI proxy like
> anonymouse.de ... TOR works fine too.

See #2 – this is not about how easy or hard it is to get around Google self-censorship. Most censorship can be worked around – that doesn't make censorship any better, or less noteworthy.

> 6) BBC is the only media censored, since CNN,
> IHT, ABC, Lemonde, and any other famous newspaper
> work fine.

No. Google News itself agreed to remove several gov't critical news sites in China:
http://googleblog.blogspot.com/2004/09/china-google-news-and-source-inclusion.html

> 7) yeah, DNS and filtering changes from
> District to District, and from Region to Region so
> we can't talk of a global unified filtering system.

Yes, you're right.

> 8) so woah there's also some of you wanting to write
> an essay on the crap you wrote.. you guys must
> be brainwashed.

You're making things too easy for yourself. Your basic point is what... that the censorship can be worked around, therefore everyone who notes it must be "brainwashed"?

> come here and see the reality, kids.

Thanks, we have self-censored Google in Germany, too. And I still think German censorship is noteworthy, even though it's trivial to route around it.