4 ms·

Web spam 2.0: 1) Take the entire works of several popular content creators in a given field, complete with links out to articles etc. 2) Concatenate them into

by dools 11y ago

Web spam 2.0: 1) Take the entire works of several popular content creators in a given field, complete with links out to articles etc. 2) Concatenate them into a single file 3) Train this thing to generate new articles 4) Create a map of popular articles that other people have written, to articles you have written on similar topics 5) Replace the originals with your articles 6) Publish millions of articles that can't be detected as spam automatically by Google It's like bot wars: Spammers can train their robots to try and defeat Google's robots.

stefs 11y ago
well, i don't see how they - the spammers - would fake google's valuation system of valuing incoming links from valuable sources. it's not like many valuable sites outside this relatively insular system would link to those generated nonsense pages. that'd practically create an insular babblenet that could be relatively easily identified. i mean, it's not like that's exactly what's happening right now.
- dools 11y ago
  Okay so in the system I'm hypothesising, I pick a topic -- say content marketing. I go to Neil Patel's and KissMetrics blog and get all their articles on content marketing, and train this thingy with them. I then buy, say, 1,000 domains. Doesn't matter what they are -- Or I buy 100 domains and setup 300 tumblr blogs, and 300 blogger blogs and 300 wordpress.com blogs. Now I drip feed content to each of those blogs, but instead of linking to the articles on content marketing that kissmetrics and neil patel originally reference, I link to articles I have created instead. How can Google tell the difference between a tonne of nobody bloggers link to Neil Patel's articles, and my bots linking to my articles? The fact is that if you blog on niche topics, with good article titles reflecting low competition long tail keywords, you'll get some traffic from Google pretty easily -- how can Google possible tell that links are coming from shitty bot generated pages versus from a tonne of obscure bloggers with virtually no audiences (of which there are thousands)? The way they can tell the difference is Panda (or Penguin? I think it's Panda ... ) so as long as your pet robot can learn from Neil Patel and Kissmetrics well enough to produce content that cannot be penalised by Panda, and so long as you don't do it stupidly by like, having the same anchor text for all the articles and doing 1,000 articles overnight and actually phase it in so that it looks as though you're getting some reasonable organic spread, you'll be able to game Google's rankings pretty reliably for your real articles that you're trying to promote, and get higher volumes of traffic to those articles than you would be able to by just focusing on niche, long tail articles (for example because you'd be able to get on page #1 or in the top 5 for much higher volume keywords). You would then get shares etc. for your actual content -- just because those "spam farms" don't have social shares or backlinks from PR6 blogs doesn't mean Google completely disregards them, just means that you need a lot more of them to make the same impact as lots of shares/backlinks from PR6 blogs. This strategy is old, and was killed by Panda, but if you could beat Panda using a RNN then this would work again.
tluyben2 11y ago
They have been doing for years and years with Markov Chains and it works if you have content farms (100.000 urls different ipv4 ranges). Usually Google weeds them out after a while but as all is automated it works. It's tricky business as Google gets better and better but it still works and people are making a lot of money with it.

by dools 11y ago

stefs 11y ago

well, i don't see how they - the spammers - would fake google's valuation system of valuing incoming links from valuable sources. it's not like many valuable sites outside this relatively insular system would link to those generated nonsense pages. that'd practically create an insular babblenet that could be relatively easily identified. i mean, it's not like that's exactly what's happening right now.

dools 11y ago
Okay so in the system I'm hypothesising, I pick a topic -- say content marketing. I go to Neil Patel's and KissMetrics blog and get all their articles on content marketing, and train this thingy with them. I then buy, say, 1,000 domains. Doesn't matter what they are -- Or I buy 100 domains and setup 300 tumblr blogs, and 300 blogger blogs and 300 wordpress.com blogs. Now I drip feed content to each of those blogs, but instead of linking to the articles on content marketing that kissmetrics and neil patel originally reference, I link to articles I have created instead. How can Google tell the difference between a tonne of nobody bloggers link to Neil Patel's articles, and my bots linking to my articles? The fact is that if you blog on niche topics, with good article titles reflecting low competition long tail keywords, you'll get some traffic from Google pretty easily -- how can Google possible tell that links are coming from shitty bot generated pages versus from a tonne of obscure bloggers with virtually no audiences (of which there are thousands)? The way they can tell the difference is Panda (or Penguin? I think it's Panda ... ) so as long as your pet robot can learn from Neil Patel and Kissmetrics well enough to produce content that cannot be penalised by Panda, and so long as you don't do it stupidly by like, having the same anchor text for all the articles and doing 1,000 articles overnight and actually phase it in so that it looks as though you're getting some reasonable organic spread, you'll be able to game Google's rankings pretty reliably for your real articles that you're trying to promote, and get higher volumes of traffic to those articles than you would be able to by just focusing on niche, long tail articles (for example because you'd be able to get on page #1 or in the top 5 for much higher volume keywords). You would then get shares etc. for your actual content -- just because those "spam farms" don't have social shares or backlinks from PR6 blogs doesn't mean Google completely disregards them, just means that you need a lot more of them to make the same impact as lots of shares/backlinks from PR6 blogs. This strategy is old, and was killed by Panda, but if you could beat Panda using a RNN then this would work again.

tluyben2 11y ago

They have been doing for years and years with Markov Chains and it works if you have content farms (100.000 urls different ipv4 ranges). Usually Google weeds them out after a while but as all is automated it works. It's tricky business as Google gets better and better but it still works and people are making a lot of money with it.