SEO

Six Degrees of SEO Separation : Can We Reverse Engineer TrustRank and Hilltop?

Imagine playing the “Six Degrees of Separation” game, but with websites. The closer you are to having a link from Maclean’s magazine or Time, the better.

The ideal would be to actually be Maclean’s or Time, obviously.

But short of that, getting a link from either site is excellent (1 degree of separation). Getting a link from a site that was linked from Time is pretty good (2 degrees). Getting a link from a site that was linked from a site that was linked to by Time is good (3 degrees). A link from a site, which was linked from a site, which was linked from a site, which was linked from Time is OK…

TrustRank and HillTop are two search engine algorithms that are based on “seed set” of trusted sites. The closer you are to having a link (or several) from the trusted sites in the seed set, the more trust you have.

More trust means higher rankings. What if we could measure the trust our site has, or that of our competitors? Here’s an idea I just had to do exactly that.

Ironically, one of the most competitive keywords to rank for online is probably not something anyone has intentionally tried to rank for. I’m referring to “click here.” Those huge sites that have millions of links, many of which use “click here” as anchor text, end up ranking for that phrase.

And while the correlation isn’t perfect, the sites that rank for click here tend to be pretty well trusted authorities.

  • So perhaps by Googling “click here,” we can form a “seed set” of trusted sites?
  • Once you have that seed set, you can come up with search keywords in Yahoo Site Explorer that let you look for links from the likes of Time.com etc. This will show you any first degree links passing a lot of trust to the recipient.
  • If you build a scraper and get clever with it, you can even dig through some of those linking sites’ backlinks, and find 2nd degree trust relationships.
  • And while the use/benefit of this data may not have been initially evident, anyone using SEO for Firefox can look for .edu or .gov sites in a given site’s backlinks. Typically, these are pretty well trusted links too.

So SEJ readers, what do you think? Is this a valuable tip for finding out how much trust your competitors have? Is it too imprecise, or is an approximation like this better than nothing? Also, here’s a ping to @SEOmoz – Is this how your mozTrust metric works?

Gab Goldenberg writes on advanced seo for his blog, often sharing new techniques and ideas.

Comments are closed.

14 thoughts on “Six Degrees of SEO Separation : Can We Reverse Engineer TrustRank and Hilltop?

  1. A lot of the Web sites ranking for “click here” in Google seem to be Web sites that offer a service or product for download (especially ones pertinent to Web site experience and have been for some time).

    For the downloads, click here could be one of the most common anchor text used (as it is one of the most common on the Web).

    – Acrobat Reader
    – Windows Media Player
    – Quick Time
    – Java Software

    A quick example: Going along with trusted Web sites, 3 out of 4 of those Web sites have links on Nasa.gov.

  2. In my research, I’ve found that the number of inbound .edu links to a domain tends to be more important than the number of inbound .gov links. Combining the two adds minor, although questionably significant importance.

  3. mozTrust does work on this principle, but like PageRank, it’s an iterative algorithm. The seed set we use, however, was based on the advice of some former search engineering folks and academic papers, rather than using a single query (or a few).

    I have to say that while we like mT today, we’re of the belief that we can significantly improve its ability to predict trustworthiness vs. spam with modifications in the future. It’s best right now when compared against mozRank – if mR is high but mT is low, generally there’s a disconnect that means spam or manipulation.

    More on mozTrust here – http://www.seomoz.org/linkscape/help/metrics#moztrust

  4. @Brandon & @Sean – Thanks for your insights guys. I agree that it is skewed, which is a problem. And I’m pleased to hear that your .edu and .gov experiences reflect what I had kinda expected.

    @Michael – I’d love to here more from you as to what problem(s) you have with this technique. I’m sure someone knowledgeable like you has some valuable constructive criticism to make :).

    @Rand thanks for clarifying! I’d love to hear more about those conversations the next time we cross paths in person :).

  5. Obviously it is all about trust. We use Linkscape a lot for this and very glad to hear Rand is going to improve it as we see it has some deficiencies.

  6. Can you provide a link to the TrustRank paper/algorithm written by a search engine rep? Links to both the Hilltop paper and Trustrank link to SEO columns. Thanks!

  7. A site… is a site .gov and .edu have no more value in authority terms than any other site due to TLD. It is myth that these are “special” beyond the fact that the research done by Universities is often cited in other authority documents. So, IMO, if you want to prove that it is easily skewed by the document set. I have a site that is over a decade old and has .edu links. It had more but… people remove links when info is very old.;-)
    I’ve never seen this so called automatic authority. Despite having links that are over 12 years old and worked with sites that had lots of .edu.

    IMO, the value of the authority of .edu and .gov in PageRank terms is the value the page has in PR Terms. Whether the green pixels or another value only the eggheads on the GooglePlex would know.

    Gab, your article outlined a good way to gather intelligence I’m only negative toward the impression given that .edu and .gov have any added value based solely on “TLD”. Sorry… gotta’ call BS on that. One should watch who and where they get their information… it goes a long way to perpetuating the myths. ;-)

  8. I think a lot of people missed the point on this. Gab, I think this is a brilliant way to explain trust to the non-SEO as well as to approach link building.

    A good exercise: use the “~” query for your top 10 or so short-tail keywords. Use the sites that show up in the top 1 or 2 as your Kevin Bacons and then go start building links.

  9. Terry,

    I don’t think anyone here is implying that .edu and .gov links carry any extra authority based on their TLD. Even if it were true, it would be extremely difficult to prove.

    What we do know is that on a macro scale .edu links tend to carry more trust/authority in comparison to an “average” link with a different TLD. The correlations are pretty irrefutable regarding this (not to be confused with causation).

    As to why this is the case, I can only speculate. My guess is that .edu domains are generally considered trustworthy for more reasons than just the TLD. Whether the TLD is even a factor at all is certainly up for debate.

  10. @Terry – thanks for the links and constructive criticism.

    @Sean – Thanks for clarifying what I was getting at. I didn’t mean to say that the TLD alone carries more weight – Matt Cutts busted that myth a while back – but that on the whole, they tend to be linked by trustworthy places and link to trustworthy places. Plus I imagine that the Ivy League has a few sites in the seed set…

    @Vin Neat tip! I’ll be sure to use that next time around :).

  11. I like this post and I’ve tried to research how Moz/Trust Rank is formed, however to no avail.

    How do you propose using Yahoo site explorer to find outbound links from a site? I know it can be done with Bing, but I’m unclear on Yahoo.