Technically speaking three fields in the squid log records are checked for this: client ip address, referer header and user agent header. A request can originate from an ip address which has been registered by Google and/or it can carry a referer tag that tells us a user clicked a link on a Google results page and/or it can carry an agent string that mentions a Google application which can reasonably be assumed to be genuinely Google's. See bottom of page for further details.

In total Google was somehow involved in 48.1% of daily external page^* requests

Google referred to our sites, through its services including search, maps, and Google Earth, 93,880,000 page views per day, representing 40.0% of our external page requests.

Including all of its different search crawlers and services hosted on its servers, Google itself requested another 19,198,000 page pages per day, representing 8.2% of our external page requests.

* = mime type text/html only

IP ranges: known ip ranges for Google are 64.233.[160.0-191.255], 66.249.[64.0-95.255], 66.102.[0.0-15.255], 72.14.[192.0-255.255],
74.125.[0.0-255.255], 209.085.[128.0-255.255], 216.239.[32.0-63.255] and a few minor other subranges

Agents: as for genuine agent strings: too many crawlers indentify themselves as 'GoogleBot' to take this at face value. They are accepted as genuine Google crawler requests only when the ip address matches a known range (see above). Other records that mention GoogleBot are counted as GoogleBot? (question mark, as this may include partners, like DoCoMo). However when the agent string mentions Google Desktop or Google Earth this is always accepted

Service: the service name is based on the agent string (plus for GoogleBot check for ip address, see above), if this is inconclusive it is based on the referer string.

Here is detailed breakdown per service of indicators that pointed to Google (total ≥ 3)
 

Service	Total	Originating from Google ip address	Referer mentions Google url	Agent mentions Google service
Desktop	2,241	-	-	Y
Earth	819	-	-	Y
FeedFetcher	350	Y	-	Y
GoogleBot	19,527	Y	-	Y
GoogleBot?	1,990	-	-	Y
Image search	3,118	-	Y	-
Mail	63	-	Y	-
Maps	191	-	Y	-
Other	52	-	-	Y
Other	10,164	-	Y	-
Other	5	-	Y	Y
Other	113	Y	-	-
Other	78	Y	-	Y
Toolbar	771	-	-	Y
Toolbar	154	-	Y	-
Toolbar	48	-	Y	Y
Translate	29	-	Y	-
Translate	131	Y	-	Y
Translate	256	Y	Y	Y
Web search	91,009	-	Y	-
Web search	27	Y	Y	-
Wireless	11	-	-	Y
Wireless	2,820	Y	-	Y
Wireless	5	Y	Y	Y

Top Level Domain 'undefined': requests with top level domain 'undefined' are nearly all requests from anonymous ip addresses (crawler and other services)

Note: averages below 1 are always rounded up to 1