Via SeoMoz I found the dutch reporter Henk van Ess posting pictures and movies from the Google Eval Lab that uses a rather low-end user interface for letting humans review sites that TrustRank or other machine techniques spit out for getting manual seed page reviews ratings.


As it seems to me this is a perfect proof that Google is using the pattern matching, fractal filter way to improve their results quality.

If you read the full TrustRank paper from the standford.edu database group then you will have an easy time to understand that 10 or 20 manual reviews can already weed out 1000s of "bad pages" from the Google SERPs by their escalating bad Trustrank

They just have to make sure they create a leverage with their algos – so that 1 human review results in ranking effects for many similar page/link patterns… no matter if it's 100 or 100,000 – the goal is same. Get most vote out of the time that one human (being paid $10-20 per hour ) spends with a review.

Of course there's no evidence that TrustRank is already implemented just as this paper details it – but it doesn't matter at all! If you take a look at all the papers published by Google's PHDs you get an idea that they won't have a hard time to have other or modified algorithms in their actual designs that use the same method: human quality reviews as signals for SERP ranking

PS: From the screenshots of eval.google.com page you can see that they have NDA non-disclosure-agreements setup with the reviewers… and I am pretty sure that one guy (the posting one or a friend) will soon get in trouble with Google – the adsense ads are already disabled there….

anyway – thanks for this Henk!

I found some other sources explaining the pretty technical TrustRank to the non-tech/non-algo user:

Aaron from Seobook
Robin Good from MasterNewMedia

With these it should be fairly clear what TrustRank is about …