It's quite simple. Every browser making a request to a website identifies itself
with a special string, the User-Agent header.
User-Agent strings are inside every web request from your browser. They could
be something like this:
User-Agent: Mozilla/5.0 (...) Gecko/2009032608 Firefox/3.0.8
User-Agent: Mozilla/5.0 (...) ... Safari/525.27.1
User-Agent: Mozilla/4.0 (...; MSIE 8.0; ... Trident/4.0)
Search engines have softwares that browse the web to read and index all
the webpages they can find. Those special softwares are called crawlers, robots or bot.
Each search engine has its own crawler, and it's also identified by a User-Agent string:
User-Agent: Mozilla/5.0 (compatible; Yahoo! Slurp; ...)
User-Agent: GoogleBot (...)
This site simply makes two requests. The first request has a RoChk special User-Agent, while
the second request "fakes" being the Google crawler.
Then, it compares the results and shows the differences, if any.
If you have a Linux, Mac or another system with curl installed you can do it for yourself:
curl --no-sessionid --user-agent "GoogleBot" http://example.org/
I know, it's not nice to fake being another software by faking a User-Agent string, but since this is just a test tool, it could be considered a legit usage.