According to The Sunday Times newspaper ("Revealed: the environmental impact of Google searches", by Jonathan Leake and Richard Woods, January 11, 2009), a typical Google search generates about 7g of CO2, about half the amount from boiling a kettle (15g). This seems a very high figure and is supposedly based on research submitted to IEEE by Alex Wissner-Gross of Harvard University. A further claim was that viewing a simple web page generates about 0.02g of CO2 per second, rising to 0.2g of CO2 a second for web pages with complex images, animations or videos. This also does not appear plausible. Further Wissner-Gross is supposed to have set up the website www.CO2stats.com. This web site is selling a service to certify carbon emissions from web sites, which makes me wonder if this is all a hoax or a scam. Harvard University list a Alex Wissner-Gross as a Postdoctoral Researcher in their Sensor Networks Lab. However, his research is supposed to be about getting computers to monitor and manage their own environmental footprints, which does not appear to have anything to do with the carbon emission of web pages.
ps: I calculated a page of a reasonably designed PDF document would produce about 1 g of CO2 a year. But this was a back-of-the-envelope calculation, not something I would consider submitting to a refereed journal, charge people money to calculate for them or claim a form of certification.