Yesterday, I wrote a post about how Google could be using the temporal patterns of paid text link buying to detect paid links. I wanted to add more examples of how Google might algorithmically detect paid text links based on Web site profiles.
I don’t think that Google looks at any one factor when trying to determine whether a site is buying or selling text links. Any of the factors by themselves would not trigger anything in the algorithm. But Google has probably mined the massive amounts of data in their Web index and determined something like “98.5% of the sites that have a combination of factors x, y, and z are buying or selling text links“.
Every time the algorithm changes it causes some collateral damage, so I think Google can live with that fact that some innocent sites will be penalized during the text-link witch hunt.
Here are some qualities that are common among most sites that buy or sell text links. Some of them might make up the factors that Google uses to determine a Web site’s profile:
- There are closely grouped sets of sitewide external links
- Each of those linked to sites has many other sites linking to it with the same profile — closely-grouped sets of sitewide external links.
(Click on the illustration below.)
Once a certain number of interlinks in this pattern are found, it could set off a flag. Both parties are guilty by association. - The sites that sell the links often have affiliate links to popular text-link-selling or review-selling services.
- Text links are often bought in 1-month blocks of time, for example 1 month, 5 months, 12 months. That could be machine-detectable.
- …and there are probably other factors that only a search engine could discover though data mining their index…
Here is a quick diagram of a common pattern that appears with sites that buy and sell text links. (Click on the image to view it.)
Below is another extended view of the same illustration. One of the sites that sells links is highlighted in blue. It links to link-buying sites from a group of sitewide external text-links. Each of the sites that are linked to from that set of text-links also has many inbound links from sites that are linking in the same way that the blue site is linking. This kind of pattern is machine readable, especially when combined with other clues.
(Click on the image for the full view.)
Paying attention to your Web site’s overall linking profile might help you avoid mistakenly getting penalized by Google and becoming part of the hypothetical 1.5% collateral damage from this algorithm update.
6 Comments
thanks really nice article. now i know why my blog (http://www.visitformoney.blogspot.com) is out of google search
Eliena Andrews
Interesting take on the algo - you’re probably not far off although I think the point of linking to “link selling” websites is doubtful.
I don’t think that Google will penalize you if you have an affiliate link to a text-link-selling Web site. But when found in combination with other certain factors it may be one of the many factors that can help profile a site that sells text links.
Take a look through the inventory of the link selling services’ and notice the high number of “make money online” blogs that are full of paid reviews and text links. Many of them have their affiliate links to text-link sellers right next to a block of paid text links.
It would be hard to know what the exact combination of factors are that Google would use to identify link sellers/buyers without having access to the data in their index.
To detect anything, like you said, you need to work with a set of signals instead of relying on one signal. A site with an aff link to TLA, for example, has a high probability of selling links, though that signal alone would generate false-positives.
A site with 100% of backlinks pointed at the root is probably owned by a link manipulator, but again, this signal alone would identify good sites like apple.com or ibm.com as link manipulators.
You combine all the signals - then you start to get an idea of the character of a webmaster - which is really all Google needs to know when deciding how to interpret a shady link.
One detection technique that people seem to be forgetting is this:
1. Look at all the links pointing at xyz.com.
2. Are any of those links obviously paid for? (e.g. listed under “Sponsored Links” or linked from several completely off-topic sites - wedding link on seroundtable, for example)
3. If there’s even one obviously paid links pointing at a site, that categorizes a webmaster as a link buyer.
4. When flipping a coin (e.g. when Google finds a contextual link with laser-targeted anchor text pointing at a site), Google can use that fact to its advantage.
Also, link sellers can get very sloppy. It’s a pain in the butt to review a site for approval; sometimes you just want to approve everything to pad your monthly paid link income. So:
1. Look at all outbound links from site A.
2. What percentage of those links point to sites that are categorized as link buyers? If the percentage is high (more than a few links point to link buyers), then the chance of those links being paid links is high.
3. Look at off-topic links to sites with low authority. An editorial link tends to point to sites with high authority, though not always.
4. Investigate the link profile of those off-topic sites. Are they link buyers? If yes, then the chance of these off-topic links being paid is high.
It’s a guessing game, trial by link analysis, and I’d expect the accuracy rate to never reach 100%, but I also expect Google to err on the side of quality search results (i.e. not on the side of innocent until proven guilty).
I think that everyone is getting sloppy. Many (most?) buyers don’t check where their links are ending up.
Sloppy link buyer: “Here is some money. You have approval to buy me any links on this topic with at least a PageRank of x“.
I wouldn’t be surprised if they do that. Google needs to make sure that they let people konw that they are being penalized, e.g., in the Google Webmaster Tools. I doubt Google will tell people though.
I totally resonate with the things you have mentioned. First time it was a guess, but as the research is going on, it’s confirming that paid links are the root cause for most of the drop.
I started two of my blogs few months back and
My one blog got PR 4
and another got PR 3
For the time being I am happy.