Ex-spam-team employee on SEO

[smithaa02]

Interesting article from one of Matt Cutts former employees talking about SEO:

[jamesnorquay.com...]

[SevenCubed]

A bit of an interesting article. The nugget in this might not be what he's actually saying in the article but the link to his own website. Guys reverse engineer it! You can be sure he has the inside scoop on the ideal way to present a website. And I don't just mean the obvious surface presentation and words but also what's under the hood. Finally some useful insight maybe :)

[themaninthejar]

"so don’t put more than 2 commercial keywords in your titles or Google will frown upon it"

Per page title or per website?

[gehrlekrona]

The most intriguing thing he says is about what they track about you, your domain registrations etc so you are ranked now, not only each web site. Very interesting...

[crobb305]

Yeah, I tested his program, and it found dozens of "suspiciously similar" sites (98% similar) because those websites supposedly used the same UPS Store (which is perfectly reasonable for website owners who don't want their home address on whois). However, I checked the actual Whois for those sites and the UPS Store address was not shown in the data, except one. The other sites were in different states! So I have no idea where that information is being made up from. If Google draws such conclusions in their quest to conquer spam, then there will be a lot of false positives.

so don’t put more than 2 commercial keywords in your titles or Google will frown upon it

Hmm, I see a lot of well-ranked sites in competitive sectors with 3 or 4 commercial keywords separated by a pipe.

[smithaa02]

On the subject of the article, the jaw-dropping thing IMO was the commercial phrase reference. I doubt they're only looking at titles... Google has the adwords data to know what phrases are valuable and aren't. I suspect EVERYTHING (titles/anchor text/body/h1's/meta description/etc) is checked to see if the occurrence of commercial to non commercial phrases is too high statistically (something I've long suspected). No wonder about/contact/privacy policies/etc... are so important...they water down your commercial keyword density.

If you haven't checked out netcomber.com yet for your favorite site, be sure to do so. If HE is using grouping data (like analytics/whois/ips/etc...) to determine groups and valid backlinks...then there is a good chance google is too.

[edited by: tedster at 5:00 am (utc) on Oct 17, 2012]

[TypicalSurfer]

I checked the actual Whois for those sites and the UPS Store address was not shown in the data, except one. The other sites were in different states! So I have no idea where that information is being made up from

Google is an ICANN-accredited registrar even though they do not offer registration services so they have access to whois data that you don't (original owner, domain history, etc.). I always thought that ICAAN did a sleazy deal with google in awarding them registrar status knowing full well that g only wanted the data for internal reasons. I believe ICAAN broke and are still breaking their own rules in that regard.

[edited by: tedster at 5:02 am (utc) on Oct 17, 2012]

[crobb305]

Google is an ICANN-accredited registrar even though they do not offer registration services so they have access to whois data that you don't (original owner, domain history, etc.). I always thought that ICAAN did a sleazy deal with google in awarding them registrar status knowing full well that g only wanted the data for internal reasons. I believe they broke and are still breaking their own rules in that regard.

Well, supposing there are other businesses with websites in the same town that use the same UPS Store, creating some level of "similarity" (or in his terms, "suspiciously similar"), then we could still see a lot of false positives if one or two of those sites are, in deed, spammy. I thought 98% similar was a bit extreme just because of an overlapping Whois element (they did not share any other characteristics, e.g., Adsense accounts or web host).

Another important point he made in the article that I think is important, and may apply in my case of OOP is:
"any type of scraped, synonymised or obviously poorly written text would be a clear spam signal."

I don't scrape and I write well; but, I have had a tendency to use synonyms. I may have to rethink that strategy.

[edited by: crobb305 at 5:20 pm (utc) on Oct 16, 2012]

[deadsea]

On the pagerank front, I had long suspected that they started discounting high pagerank links. From this quote: "Not only this but take PR for example, getting a link from a high PR page used to always be valuable, today it’s more the relevance of the site’s theme in regards to yours, relevance is the new PR" it sounds like a high page rank link might be great if the linking site is on-topic.

"if you want to please Google with your SEO, then forget about SEO" is horrible advice. If I were not worried about SEO, I wouldn't let robots crawl my site. I wouldn't write page titles that users barely see. I would a lot more flash and ajax to make the presentation flashier. The advice to be passionate and create good content is good, but you need to worry about search engines if you want visitors.

"We use signs like ... coding style to determine ownership clusters. ... Search engines always used this data in their fight against spam" Coding style is pretty advanced stuff. If you are trying to hide the fact that you own a site, I guess it might be worth using a standard CMS system (different than the one you usually use).

[smithaa02]

Deadsea...agreed on the pagerank front. In the article that Tedster posted from the 'other ex-googler' :P...that engineer made an interesting point about emphasizing links contained the keyword referenced. So if you want to rank for couches...your referral linking pages should have the word couch or some synonym on that page or the site. Getting a link from a website about say 'pets' that doesn't feature the word couch or variants, even if it has a high PR, probably won't help.

I also think the opposite is true. It HELPS you to link to sites that are in your keyword theme...and hurts to link to sites that are out of your theme.

[brotherhood of LAN]

>We use signs like ... coding style

And there would be a lot of signals in that group, from tabulation, ordering of tags, nesting of tags, ids, classnames, preferred frameworks....

[HuskyPup]

Hmmm.?.?.?

First of all, choose your domain name wisely, having a good URL can give you a head start in the race. Good domains are still expensive and for a good reason.

So exactly where does this leave the EMD/PMD update?

[johnhh]

We use signs like ... coding style

Every website designer has a 'style', so if G doesnt like his style that puts all his clients websites at risk ?

[TypicalSurfer]

The big picture takeaway for me is that they are assigned to "markets" which really means money. Most of what they are deeming spam-fighting could very well be called revenue protection, there is only a difference in the semantics used. They even have us trained to call revenue encroachers things like "spammers", "black hat SEO", etc.,... the "dark side of the internet". LOL

[driller41]

coding style to determine ownership clusters

Link that to IP address at the least and it must be easy to pick up multiple sites hosted on one server.

[bwnbwn]

the dude has put together an intresting website. I did a seach and it shows Sites suspiciously similar to x looking up one of the similar domains is not live and the domain owner is hidden, so how is website able to connect them?

I did one of my own and it nailed mine but I don't hide my info so mine is easy.

[johnhh]

@bwnbwn
Did one of mine as well, Sites suspiciously similar to x came up with a shop down the road, same start of postcode (zipcode) nothing to do with me. Look too noisy to me. Looks like 'nofollow' may actually flag up something.

[TypicalSurfer]

the dude has put together an interesting website.

It definitely has that creepy factor going on. I was looking around for a UA to ban but I suppose he may be above all that. Also tried his domain to see what HE may be hiding and got a "Mind your own Business!" message. He may be terminally infected with hubris.

[nomis5]

Supremely interesting insight - it's now on my desktop.

The lesson i take from it is that big sites, run white hat, will win over small sites. A small site can so easily and innocently run into trouble. A large white hat site will smooth out any problem areas as far as G is concerned.

The problem for the committed small time guys like me who have only one "big time" site is how to protect yourself from that big time site falling foul of G. I did have a backup plan but the EMD update crucified that.

I have a couple of other sites which are OK but big sites seem to be the name of the game nowadays. This article makes me feel very, very vulnerable.

[1script]

Well, Andre has definitely taken his "Trust no one" Google Spam Team approach to his new project, netcomber.com !

I have an issue with the entire premise of "Sites suspiciously similar to ... example.com" Sure, it's finding most of my sites (perhaps all - you need to register to see for sure and I didn't). But there is nothing suspicious about it (no effort to conceal it) nor they are actually similar! What makes two sites similar if all they have in common is a small part of a Whois record? And there is A TON of false positives based on similarities of WHOIS protection service return emails.

In other words: well, OK, there's a way to find SOME relationships between given sites. But not all of them should have spam-related implications! If this is the sort of an approach Google Spam Team is using, it's outrageously irresponsible in general, and even on the practical level I imagine the false positives create a lot of useless extra work.

[bwnbwn]

johnhh intresting he connected the zips off the sites or register?

[crobb305]

I have an issue with the entire premise of "Sites suspiciously similar to ... example.com" Sure, it's finding most of my sites (perhaps all - you need to register to see for sure and I didn't). But there is nothing suspicious about it (no effort to conceal it) nor they are actually similar! What makes two sites similar if all they have in common is a small part of a Whois record? And there is A TON of false positives based on similarities of WHOIS protection service return emails.

In other words: well, OK, there's a way to find SOME relationships between given sites. But not all of them should have spam-related implications! If this is the sort of an approach Google Spam Team is using, it's outrageously irresponsible in general, and even on the practical level I imagine the false positives create a lot of useless extra work.

Exactly. Overlapping Whois? It's a coincidental similarity at best, certainly not suspicious. This is something I'd expect from a college computer science capstone project. It has no "spam" implications whatsoever, and surely the experienced staff at Google know better. As for similar Adsense codes, they already know who those webmasters are.

Every business using the same UPS Store location is shown as "suspiciously similar". Ridiculous.

[bwnbwn]

Every business using the same UPS Store location is shown as "suspiciously similar". Ridiculous.

Maybe not think about it, the search connects them now the code can be examined if the code is very close to being the same it could lead to another closer examination of the sites. I have 20 websites using the same ups shipping location they are all registered in different names different servers etc. yet he can connect them from a ups shipping location. Now I can look at the source code and see if the same programmer coded them. Not many code the same we all have our quirks. I can see this could be a real problem for Google's team as to why the delete of the article.

[nickreynolds]

This article is actually one of the most helpful things I've read in a long time. Ive long subscribed to the view that some innocent sites get hit because they are "guilty by association" (for instance if they have a number of things in common with a number of other sites that are perceived as being spammy, such as ip address, common adsense code, common analytics code etc).

[edited by: nickreynolds at 10:09 pm (utc) on Oct 16, 2012]

[johnhh]

@bwnbwn
off the site , we are legit uk company, no relation to shop via register or ip.

[bwnbwn]

nick my exact thought as well. I am looking at this new read very very hard and doing some research on my own. If I can connect a website through a link profile well Google has already done that.

john was connected by a zip code others usp shipping others by register. Think about it once they are connected then a deep look maybe another type of tool Google developed that ices the cake.

[Nostalgic Dave]

If you take a look at the source code of the home page on Mr. Weyher's website, you'll see that he's done virtually no on-page SEO at all! The page description is just one word (the name of the site), and the Google Analytics tracking code has the deafault UAxxxx still in place... no real tracking code! He's either sloppy, or this is telling us something.

[phill george]

No analytics = minimise your footprint

[smithaa02]

Phill...who knows...maybe not having GA and say private whois is a sign that you are hiding things...and this factors against you :P

On the subject of GA, there was a previous thread on the subject of google using this as a tool to track down link builders:

[webmasterworld.com...]

Basically, Chris seems to have proof that it does happen.

[topr8]

>>Link that to IP address at the least and it must be easy to pick up multiple sites hosted on one server.

with a simple tracert you can pin sites you suspect to be from the same server no problem, i always wondered why some guys were so desperate to have different c class ip addresses but used the same server - like it in any way hid what you were up to! (i'd say it earmarked you as a spammer! always have thought that)