What is Google doing to our language?

The other day, I attended a course in the lab on concordances and how these can be used for linguistic research and in language education. Jon, who was giving the course, pointed out that the most commonly used concordance tool today probably is Google, and we had an interesting discussion about the consequences of this.

I am sure many of you use Google to check different constructions and compare frequencies; at least this behavior seems to be common among those of us who do not have English as our native language. But how reliable are these results? In some cases the frequency numbers differ so greatly that there is no question that one construction is incorrect. However, sometimes the differences are not as great, and a more thorough analysis of the results is needed. For example, one has to evaluate some of the sources for the hits found (for instance, if you only get a small sample and the majority are from .se addresses, it is likely that we are only dealing with a common Swenglish expression).

One advantage I see with using Google instead of traditional corpora for concordance is that Google captures language at it is used today. We may well find examples of constructions previously considered as incorrect which now appearto be common, and this might lead us to accept that language is constantly changing. Maybe to sometimes split infinitives is not that terrible a crime after all?