The Big G is to access Google (and Bing by special request only) and does not work with any other sites.
Your queries to the Big G server are sent by HTTP, whereas your actual queries to Google are sent via HTTPS
URL's need to be as organic as possible and you do not use unnatural query constructs, e.g. num=100 , or the gl and uule tags (which are incompatible with The Big G). Google is now very sensitive to manipulation of the URL when mass scraping.

Queries are throttled to a maximum of 1,000 queries per minute.
Close the TCP/IP connection after each unique keyword query completes otherwise this will severely limit the capabilities of The Big G.
Cookies can be used but must be disposed of properly after each unique keyword request completes.

You use random, up-to-date user agents but retain the same user agent for each SERP for each keyword. For example retain user agent A for pages 1, 2, 3... of your first keyword query. Retain user agent B for pages 1, 2, 3... of your next keyword query.