Scary to think that a major search engine really uses regular expressions heavily. Regexprs are great for quick scripts, but one would expect that in major production applications better and higher level parsing algorithms would be used. It must be a nightmare to debug if you have a lot of reg-exprs interacting in a large code base.
If the language you're trying to parse happens to be regular, then regular expressions are the perfect tool for the job. They're simple, reliable, and impossibly fast.
When you think about it, a search engine is mainly about finding key words in a huge pile of text, without caring much about the structure of the language the text is written in. That use case is totally specifiable by a regular language.
