Different C Claass IP for every site?

[Holis]

Is this even worth messing with? If so do you know of any good hosts that offer it?

[Tim]

Skynet hosting offer this. However, you're still going to have the same nameservers.

[Michael Martinez]

There is minimal SEO value in spreading your network across multiple C-class IP addresses. This concern came out of the SEO community overinterpreting and misunderstanding two Google technical papers that were published several years ago.

The most influential paper explained the Hilltop algorithm, which has only ever been used in Google News search (although I suspect they may have recently implemented it in Google Blogsearch). Google employees have repeatedly denied the use of Hilltop in Google's Main Web Search.

The other paper described a technique that has, to my knowledge, never been implemented (or if it was it didn't last long). ON EDIT: Or perhaps they use it in Google Scholar.

Both papers associated class-C IP address blocks with "hosts", but they were extending the definition of host beyond what is usually used by search engines.

In normal SE technology, a "host" is any domain or sub-domain. A host may share its IP address with other hosts.

The research behind the two papers I mentioned above was based on very limited scope data -- news sites in the case of Hilltop and university research sites in the second case. Neither data set was representative of how the greater Web actually works.

I've learned to live with SEO community expectations of separate C-class IP addresses in many contractual relationships but it's really an unnecessary measure in most SEO processes.

I would consider it to be marginally helpful in a massive link-building campaign, but I wouldn't worry about nameservers unless you're using your own nameservers.

[Jeremy]

Thanks for this post Michael. We discussed this on our radio and the consensus was the same. It's good to hear it from an expert.

[scrawnybob]

I debated this for a while and came to same conclusion - that if you werent doing anything bad - then why hide it

also heard logic (not sure others might be able to add more)

that given that google does work on good / bad neighbourhoods - you could end up in a "bad" neighbourhood of other webmasters sites -

so logically if you have all (or most, or half) of your sites in the same host / IP / block however you look at it - neighbourhood and are responsible then ...

your effectively creating your own "good" neighbourhood ... less likely to get tarred with the same brush for other sites behaviour and over the long term creating a stronger basis for things

not sure how much water that holds but afiaks if your linking to your own sites based strictly and only on relevance - then your unlikely to get penalised - and IMHO you shouldnt

(if I can find a relevant page on another of my sites - why the hell shouldnt I openly link to it ?)

however if your linking on mass / excessively or not related links to your own sites regardless of trying to "cleverly" hide it - you would get penalised

I came to the conclusion all the IP block stuff was nonsense and only something you would need to consider if you were pushing the realms of acceptable linking ...

meaning its the linking that you'd need to rethink not the hosting etc

cheers scrawnybob

[Michael Martinez]

Google has never, to my knowledge, divulged the criteria by which it judges neighborhoods but their occasional comments lead me to conclude they are looking at linking relationships (who links to whom).

Keeping in mind that what follows is speculative, I propose the following:

You have three kinds of linking relationships with other sites:

• Reciprocal (you to them and them to you)
• One-way outbound (you to them)
• One-way inbound (them to you)

A "neighborhood" has to have more than one house in it, and the houses have to be connected to each other in some way. Here are several possible ways to define connections between Web sites:

• Tight interlinkage - They all link to each other (like a link farm) or mostly to each other.
• Common hub linkage - They are all or mostly listed in the same directories and hubs.
• Common referents - They all (or mostly) link to a core group of Web sites.
• Common on-page factors - They all use certain expressions, especially in page titles, meta descriptions, page URLs, and maybe Hx or large font expressions.
• Common anchor text - They all share inbound anchor text expressions.
• Failure to obtain trusted links - They all optimize for an expression but no trusted sites link to them.

There are surely other ways to define a neighborhood, and your definition could be refined by combining two or more of these criteria.

So let's say we have a black box that identifies neighborhoods. You have a Web site that links out to another site with a specific expression. You feed the expression to your black box and it spits out a neighborhood (assume that's a list of Web sites all deemed to be related, not necessarily through common anchor text or on-page expressions).

The site linking to a member of the neighborhood may not be a member itself -- in which case you want to know if the site is linking to a "good" or a "bad" neighborhood. Your black box may be able to do a lookup that determines if the neighborhood has been flagged as 'bad".

You end up dealing with thresholds in this kind of theory. How many "bad" sites make a neighborhood bad? How many types of behaviors make a site bad? How many outbound links to bad sites and bad neighborhoods do you have to have before your own site suffers?

Most people who fuss over this stuff (that is, who are actively linking to sites they feel may be suspect in the eyes of the search engines) tend to favor the 80/20 rule. That is, if 80% of my outbound links point to well-known, highly trusted sites, maybe I can allow 20% of my links to point to sites I'm not so sure about.

Understand that most people would not necessarily link that many sites they are know are bad. I'm just saying they don't know for sure whether the remaining 20% of destinations are trusted, penalized, or whatever.

Some people use the 90/10 rule rather than the 80/20 rule. Neither threshold has been shown to work, but it is considered to be axiomatic that the fewer bad sites you link to the less likely your site is to be deemed a bad site itself.

Of course, your site might only link out to good, highly trusted sites but it might be flagged as "bad" for other reasons.

It's not all about links. In fact, it's not about links at all -- it's about behavior, or what you do with what you have. Will your visitors be safe? Will your visitors be mislead? Will your visitors find anything of value?