Well the key is to find unique data in the source of the page ('View > Source' in Internet Explorer). Think of it this way, if the site is myspace.com and the keyword is myspace.com then if the Proxy returned an error like:

Sorry, we cannot contact myspace.com at this point in time.

You would have a problem as your keyword would still be found, and you would get an invalid result. The trick is to go into the source of the page and find a short string which wont change on you to scan for. A couple to mention are as follows:

coolNewPeople
fuseaction
et="_blank">Promote
ss="more">[more music]

As you can see, it would be very unlikely for that data to change, and for a Proxy to accidentally return it unless it managed to contact the site. The keywords I selected in the examples are all what I deem most unlikely to change, and the most unique.