martinibuster - 9:49 pm on Jun 6, 2013 (gmt 0) [edited by: martinibuster at 10:36 pm (utc) on Jun 6, 2013]
The question is how does Google identify advertorials with do-follow links (especially if the advertorial doesn't contain tell tale keywords like "sponsored post")
I speculated on a plausible way of doing that in my post above. Here it is for those who don't like to scroll up:
One way I can think of is to use data from confirmed link spam and identify sites publishing articles that link out to the websites in the group of confirmed link spammers, then penalize sites that have a high percentage of outgoing links to sites known to engage in article publishing spam.
If a certain percentage of outgoing links from a website are going to identified pay for play spammers or article directory spammers, then that website may fit the profile of a pay for play article publisher. Algorithms work with pattern recognition. So if a website fits a pattern then the website is guilty.
In case the above is difficult to understand, I'll add more details. Google has a data set of spammers who use article directories. Those have been penalized, pandalized, penguinized and whatever other -ized you want to add to the list. Let's call that a set of known article directory spammers.
Drastically simplified description of how to find guest article link sellers
You take a set of all web pages that are guest posts, guest articles, and similar wording. Weed out all web pages Google already knows about, TLA, article directories, known link seller networks, etcetera.
Now you pretty much have left over the sites that have not been penalized, as far as we know. This is the set we want to get to. How we get to that set is unimportant. What we're starting with are websites that publish articles that haven't been penalized. In the previous paragraph I outlined a drastically simplified description. For argument's sake we can say Google is dunking websites in a tub of water and if it floats it's guilty and leave it at that.
The point is Google gets to the set of sites that have not been identified and are otherwise unpenalized. One thing Google can do is to analyze outbound links and find published articles with links to sites that intersect with the set of known article spammers.
One objection can be that maybe these spammers are simply aggressive and the publisher of those spam articles did not accept money and it's all legit. If someone had raised that objection to my post I would have said, wow, that's a good objection. But no one did. So I'll raise it here.
One objection can be that maybe these spammers are simply aggressive and the publisher of those spam articles did not accept money and it's all legit. It's possible. That's called a false positive. False positives are always a possibility. That's why Google creates a sandbox index to test the algos before letting them go live. So how do you prevent a false positive from ruining the legit article publishing party?
Statistical probabilities is the answer. Statistical probability has been the leading answer to weeding out false positives for link spam fighting for years and years. I know of at least one guy on Matt's spam fighting team who has an advanced degree in that. Don't know if he's on the team anymore, but I found it interesting he was on their link spam projects. Don't you?
Statistical probabilities is how you weed out false positives. You take a sampling and find a percentage of known link spammers in the outbound links in articles. Then you hand check those sites for false positives. A quality web publisher won't be taking articles from an SEO agency in Ohio on behalf of a client then linking out to both teh SEO agency AND the client, that kind of thing, you know, the one called Reality. :) Once you find the sweet spot, you're good. That's a drastic simplification. But it illustrates the process, the way link spam fighting has been done for years and years.
Way back in the TLA heyday, when leading SEO types were recommending TLA [web.archive.org], it was (wrongly) argued that Google would not penalize TLA links because Google couldn't identify them and because TLA was "legitimate" advertising. Both arguments were self-interested nonesense and you never heard such drivel or testimonials from me. I was invited to put my name to TLA marketing, but I declined. Well, it's 2005 all over again. You can bury your head in the sand and say it's FUD because Google can't do it. OR you can think about all that Google accomplishes already and understand how easy it is for Google to do.
I personally can't remember seeing an online advertorial that didn't have a do-follow link.
? You're confusing me. Are you saying you have never seen an advertorial with a no-follow link?
[edited by: martinibuster at 10:36 pm (utc) on Jun 6, 2013]