The Hilltop algorithm was proposed by two researchers working under the auspices of the University of Toronto. One of those researchers is now at Google and Google also acquired the patent early in 2003. The algorithm overcomes problems with broad search terms which return large sets of documents that have to be ranked. It is hard for a search engine to analyze the quality of these results based on on-page factors alone, especially where the results are heavily optimized for search engines or are copies of other high ranking pages.

The Hilltop algorithm is based on a similar assumption to Google PageRank. That is the quality and quantity of inbound-links is an indication of how the page should be ranked. The key difference is that only ‘expert sources’ derived from the query terms are used when judging inbound-links. In other words Google’s original PageRank algorithm gives a global rank of the quality of a web page. Hilltop determines the quality based on the relevance to the query term.

An expert or authority page covers a certain topic based on the query text and has links to many none-affiliated pages on the same topic area. An example would be a Open Directory Project (DMOZ) or Yahoo! directory page. Search results are only considered if they have more than one inbound-link from these expert pages and those inbound-links contain anchor text that matches the target page and the query terms. If Hilltop can’t find more than one expert page it returns no results. The algorithm is oriented towards quality not quantity of results.

To score highly under Hilltop sites have to acquire inbound-links from expert sites. This can be by writing articles or providing other useful information. These can be placed on the website and possibly submitted to authority sites with a link request.

Sites are judged to be affiliated if they are on the same Internet network (using IP address information) or the same domain. This should help to exclude link farms.

The exact implementation of Hilltop is described in Bharat and Mihaila’s paper: . It is believed by some experts to form a significant part of Google’s Florida algorithm changes that penalized many heavily optimized and commercial web pages. An obvious problem is the compilation of the ‘expert pages’ and the processing necessary to match with results pages. It seems likely that the Hilltop algorithm is only run on certain popular keywords with a precompiled set of expert pages.