Yahoo's Zawodny In Paid Links & Nofollow Debate

[PhilC]

Originally Posted by joeduck

Who's in charge here anyway?

Webmasters/owners are in charge of websites, and search engines are in charge of search engines. You wouldn't think so when you read Matt's blog, and see people asking him if it's alright to this, and if it's alright to do that, on their sites.

I agree with projectphp that links are the hardest factor to manipulate. They are not difficult, but everything else is on-page/on-site and a heck of a lot easier to deal with.

As a wise and ancient sage once said (a few posts up in this thread), when Google launched their engine, they actually screwed up the natural linking of the web - the very thing that they rely on so much - and, until the large links-based engines move on to a better way of ranking pages, the natural state won't return. Now they are forced into dealing with the mess they created, and there can't really be any serious objections to them doing it, as long as they don't go too far and actually penalise sites for displaying paid links without the condom. That would be going much too far.

If they stop all OBLs from carrying weight just because a site contains paid links, it wouldn't be detrimental to the site. It would be detrimental to Google, because they would be rejecting a bunch of natural links - the type they want - but it wouldn't hurt a site. In fact it may well benefit a site by making a PageRank hog.

[orion]

Actually, I know first hand some at this forum that are into link spamming Google but is up to them to show you their true face.

Why not apply that to links...

Actually block analysis is applied to links and almost any block of information. Just do a search and you will find the block level link analysis thread.


Orion

[projectphp]

I guess I just wonder why analysis of code, in which different elements matter more than others, makes sense but disregarding links doesn't make sense...

Willing to here some justifictaion

[dannysullivan]

Danny. I haven't understood that all of a site's OBLs are discounted when there are some paid links on it. Are you sure about that? My understanding has been that only the paid link are not counted.

I'm right with David when he said:

I think Danny was saying that Google could do this, not that they always do. However history has shown that Google can single out a site and not allow it to pass any link reputation whatsoever, paid or non-paid.

That's it, exactly. It's something they could do. It's something Matt has heavily implied that they will do. But in reality, they might only do block level filtering, picking out only the paid links if they can identify them to determine if they pass PageRank.

It should be easy to test. I've had people tell me that our sites no longer pass PageRank. Someone can probably verify if this is the case for a "regular" link versus our paid links/ads in SEW Marketplace. Same too for Jeremy's site and for O'Reilly.

As I said in my article, it doesn't hurt us if that's happening. We don't sell our ads with any claim that they're going to boost you in search engines and never have. What it hurts is Google, if they are indeed not counting any of the links out of here, because they are then preventing themselves from tapping into the highly edited and authorative linking we do on a regular basis. Fair to say, the links we put into articles far, far outweight the quality of the links Google scatters all over the web in AdSense.

I don't follow why paid links are to be regarded as inherently bad.

All paid things are bad, Brian. Just understand that.

Seriously, paid links are not inherently bad. People have been and still do buy them for reasons other than search ranking. AdSense is nothing but paid links, and Google makes a lot of money off AdSense.

What Google especially, but all the search engines to some degree, worry about is that paid links can undermine the "democracy" of linking that they still heavily weight their ranking systems on. Links are votes for the best stuff, to some degree, and paid links allow people to buy votes.

That leads to:

On and all, there are many reasons for coming with a better solution for scoring relevancy. The commercial aspect is only part of a larger problem.

Exactly. Links worked well when there wasn't much awareness and trading of them. But they have increasingly become devalued for various reason.

As blogs have exploded, and they put sites into their blog rolls that appear on each page of their blogs, them's a lot of links My understanding is that the various search engines have tried to understand if they see the same link in the same place and with the same anchor text this way. If so, they may try to downweight. Instead of giving the link credit for all 100 appearances, for example, it might get counted once.

This remains speculation, but Google in particular has dropped so many hints that it might do this that I tend to believe it. It also correlates well with various Google updates we've seen, where it doesn't appear that sites are actually penalized but the links they've used to rank well have suddenly not counted for as much.

Paid link networks have also exploded, of course. And the reason this debate has a new chapter is nofollow. Suddenly, search engines can say, "look, if you're selling links and really mean us no harm, really aren't trying to sell link juice to influence us, use nofollow."

Initially, it sounds reasonable. I'm actually pushing to get nofollow added to our ads simply to reinforce that they really aren't something meant to influence search engines. Our ads are all to be restructure soon, so hopefully that will happen as part of it. Ironically, it won't get us any benefit. The ads aren't sold to push link juice, so nofollow doesn't take away from that. The change might get us in Google's good graces to pass link benefits -- but that's not really a plus to me -- that's a plus to Google. And that's assuming we have been actually blocked this way. I've heard that from third parties. It's not something I've ever heard from Google. If it is indeed happening, it underscores Google being sucky in not giving me any idea of what it is they don't like.

But using nofollow doesn't seem so reasonable when you look at all the arguments that came up in O'Reilly In Off-Topic Link Selling Debate and Matt Cutts Comments On Reputable Sites & Link Selling, where we get some gray areas about links being sold or traded in other ways.

Two good examples come to my mind. First is when Robert Scoble did his Search engine optimization, blogger style post (and see my response that he posted here ). In that, he links to his brother's site and help it rank well for "silicon valley law firm." That's not a paid link, but shouldn't it be treated differently from a "real" link. After all, a primary reason it is there is that Robert's helping out his brother. Note that I don't mean this as a slam against Robert, not at all. It's just a good example of where you could scream "nofollow, nofollow" in exactly the same way you do with paid links.

The other is something Andrew Goodman blogged about not to long ago in Help Us Build a Link Farm, Get an iPod Nano?. So if I have a contest or do a barter to get a link -- and links that might not be in clearly identified sponsored areas -- shouldn't those have to take nofollow?

And heck, I talked a bit in Googlebombing Now A "Prank" And Not Web's Opinion, Says Google about what's tolerated in linking as long as it's not paid. So all these people doing various link bombs, that's now considered a "prank" rather than link democracy in action. Yet no one is telling pranksters, "hey, please put nofollow on your links, because you're screwing with search relevancy."

So nofollow isn't as reasonable as it sounds. I still tend to lean toward saying it's something publishers who want to sell links should probably do, especially if they don't want potential public relations fallout. But a part of me feels like it's sort of going after the most blatant, red speeding sports car you see on the freeway while ignoring all the other cars going just as fast or driving more dangerously. It's not really solving the problem and kind of feels unfair.

Worse, it doesn't get at link networks that are developing where people are being approached to drop paid links within editorial content. Those are harder and less industrial to develop, but they exist now and will only grow as other link networks fall under pressure. Some people will get outed, but the search engines will never spot and discount all those links.

So...

IMHO, Google place "a strong emphasis" not on devaluing links, but correctly valuing links.

Right. It's never, ever been that all links are created equal with Google. PageRank, the popularity component, established a way to decide how much weight each link should carry. Anchor text is that other crucial component. And over time, exactly how links should be weighted and counted has continued to be reshaped.

The problem is, it's like propping up a house of cards. Now that so many are aware of the power of links, there's buying, selling, swapping, bartering, bombing -- you name it, many things being done overtly and inadvertently to influence search engines. And they keep trying to come up with new refinements to keep the system hobbling along. That's one reason why personal search is so much being examined. They need something else, because there's only so much link tweaking you can do.

Jeremy and I were talking at WMW Las Vegas about how publishers are now *changing what they write* to conform to Google's Algorithm and thus rank better. This has a lot of implications and most of them are negative for the web.

Absolutely, but to be fair, people are changing even for Jeremy's employer Yahoo. Part of the reason the issue has developed around what Jeremy is doing isn't because Google says don't buy links. Yahoo says it too. I have no reason to believe that Yahoo's search quality team is going to be that thrilled over Jeremy carrying links like this. It undermines entire what they say. Here's what Yahoo's Tim Mayer said literally last week:

Don't use link exchanges or buy links

Who is going to believe that or want to respect that when you can go over to one of Yahoo's most visible spokespeople and buy links on his personal blog? Who's going to think they're being "black hat" by doing that?

Aside from links, we have long, long changed what we do for search engines. We do it for good reasons and bad. Yep, we don't want search engines to dictate what we do. But then again, you don't want to simply be stupid and say you'll build your site out of frames, if going another way means you'll get traffic. People who want to be on television, covered in the news and so on approach those media in certain ways. Search is a medium. It comes to the table with all the baggage and issues of other media, and we're sorting through that mess.

[projectphp]

The problem is, it's like propping up a house of cards. Now that so many are aware of the power of links, there's buying, selling, swapping, bartering, bombing -- you name it, many things being done overtly and inadvertently to influence search engines.

Maybe it is my philosophical disbelief in Utopia, but I really don't see that as a problem. There comes a point at which a game is played for no higher reason or goal other than to maintain equilibrium. The tax system is one such example, and links and link pop are going the same way.

I see no real issue with links as a game of equilibrium, with each side playing the game. SEOs trying to find good linsk to buy. SEs shutting down link selling sites. To me, that is almost a noble game, and one that makes a lot of sense. Relevance isn't static, it is moving, and as long as you are fast under breaks, you can cop a bit of loss on the straights. With teh link game, each side will win a few and lose a few, but in the end it will all pretty much be wash. That is especially true of the very broad use of "links" we are using here.

Like Tax, I think SEO may be a case of staying one step ahead, and hoping the money you spend is < than the money you get back. Businesses accept accountants, why not links and SEO? For the SEs, link detection, like my favourite all time acronym, ECCM (Electronic counter-counter-measures) might simply be a game they need to play to achieve the best results.

[David Wallace]

What Google especially, but all the search engines to some degree, worry about is that paid links can undermine the "democracy" of linking that they still heavily weight their ranking systems on. Links are votes for the best stuff, to some degree, and paid links allow people to buy votes.

Obtaining links is becoming very much like our political races in the US. Many times the politicians that have the most money win the election. The ones who gain the most exposure via advertising are often those who are most prominent in people's mind. The best candidate for the job does not always win.

Search results are becoming the same way IMO. It used to be that the quality and content of a site would push it to the top but nowadays how many times do you see a site that is on top largely due to links that really has no right to be there?

Therefore I can certainly see why search engines are trying to fix this problem in the same way laws are being put into place regarding political races (i.e spending limits, term limits, etc.) all to try to make it more fair for all the candidates.

So the challenge for search engines is to try to do all of this in a fair way. Right now it is as Danny says, "it's sort of going after the most blatant, red speeding sports car you see on the freeway while ignoring all the other cars going just as fast or driving more dangerously. It's not really solving the problem and kind of feels unfair." I wholeheartedly agree. It would be nice if everyone who sells links would separate those from editorial with the "nofollow" attribute but the reality is that not everyone will.

Therefore I feel the search engines are going to have to come up with a more intelligent way to identify paid links. I don't think they will ever fully conquer this however in the same way that they still haven't figured out how to fully deal with spam.

[Chris Boggs]

OK so bear with me I am going through this chronologically to make points and ask questions, so forgive me if you answered in a later post.

Originally Posted by orion

It transpires from conversations with IR colleagues that they just brought this onto themselve and now they may need to come up with a better solution because evidently "nofollow", proposed as a noble idea to fight spam, is not the best one. Do I have a better solution? I'm not sure if they are The Solution or the best one, but I have several ideas I'm not afraid to test if you ask me.

In my view, trying now to assign relevancy weights to some links but not others because third parties are monetizing the link structure of the Web or using it to pass noisy weights is just hypocritical thinking.

If others want to monetize links, let them do it. Just come up with better models for scoring relevancy.

As usual I am pretty much right with you on a link-related matter, orion. I would certainly agree that there can be a better way. However I am confused by your last two thoughts, which I must be misinterpreting. You think it is hypocritical to now assign relevancy weights, but then say we need a better model for scoring relevancy? Should this model then be applied to only "new" links, essentially "grandfathering" established links, "nofollow" or not? Sort of like what Avaden was getting at I think?

Originally Posted by PhilC

Google isn't saying what people can and can't do concerning links. They can't do that because they have no authority, and they don't try to do it. All they say is that, if they spot bought links, presumably those that are bought for ranking purposes, they won't count them. It's not a penalty. The links just won't be of any ranking benefit.

How on earth (or in the Googleplex bio/technosphere) could a search engine accurately (even 50% of the time) determine such intent? What if the link has equal SEO and traffic value?

Originally Posted by dannysullivan

In yet other words, think about it like this. Google would prefer that paid links carry nofollow so that people can't just buy link love. However, if a site is unwilling to do this to paid links -- and Google becomes aware of the site -- then it effectively slaps nofollow down on ALL the links in the site, not just the paid ones.

This seems unacceptable. What if I am a developer who is not even aware of the "nofollow," and therefore don't use it when selling a few links. I may even be unaware of the SEO benefits I am passing-on. Then I get dinged for all the other links on my site which lead to further relevant content that I placed there for the benefit of my visitors? Wouldn't this unfairly penalize my own site's ability to rank? All this is possible, and I feel that G cannot consistently identify intent.

Originally Posted by projectphp

My $0.02.

I agree with many of your points, especially the idea of SE's packing up shop and links becoming even more valuable. BUT, when you say "links are the hardest factor so far thought of to spam," you must be ignoring the tricks being used by more advanced link-spammers that don't care about the "rules." If they can do it, anyone can, IMO, which makes it far harder to "easily spot those that stand out, and disregard large chunks of whatever."

Originally Posted by orion

Disagree. Block analysis is not aimed at ignoring anything. It is a segmentation approach to score blocks of related information.

and later

Originally Posted by orion

Actually block analysis is applied to links and almost any block of information. Just do a search and you will find the block level link analysis thread.

So can you briefly comment as to the ways block analysis can be adjusted to ignore something?

Originally Posted by I, Brian

If not, then why does Google have such a strong emphasis on wanting to devalue links? If you devalue paid links, do you really see an improvement in quality of the SERPs?

Exactly. Unless you can do this on a "block level," with a fail-safe (IMO at least 85%) system. If you block all links on a site, you are hurting its ability to prove that it leads to further relevant content. (I think I should create a shortcut key for the three words "further relevant content")

Originally Posted by joeduck

Google, through its massive search market share, is changing the way most of us write for the web, manage our links, and decide on topics. I wish people would wake up to this fact and think more about it's implications for the future of the medium.
Who's in charge here anyway?

Good point, it brings up: I and other copywriters have seen excellent results by focusing on writing and linking for the visitor, which seems to more and more parallel what Google and Yahoo! are looking for. Over-optimized sites often stick out like a sore thumb, and could possibly lose traffic at a greater rate than increasing average pages viewed or other important traffic factors. It is also harder to gain "natural" (unsolicited) links for these sites, I would venture to guess.

Originally Posted by projectphp

I stand by that statement. "Harder" is relative, and if anyone can name a harder to manipulate, not absolutely impossible just more difficult, ranking factor than links, I will willingly conceed the point

Can you manipulate relevant and consistent legitimate traffic from authority sites? I don't know I haven't tried this but it would probably be harder, in my experience. Had to take a stab at that one.

[Chris Boggs]

Originally Posted by dannysullivan

(1)...It also correlates well with various Google updates we've seen, where it doesn't appear that sites are actually penalized but the links they've used to rank well have suddenly not counted for as much.
(2)Paid link networks have also exploded, of course. And the reason this debate has a new chapter is nofollow. Suddenly, search engines can say, "look, if you're selling links and really mean us no harm, really aren't trying to sell link juice to influence us, use nofollow."
(3)The ads aren't sold to push link juice, so nofollow doesn't take away from that.
(4) In that, he links to his brother's site and help it rank well for "silicon valley law firm." That's not a paid link, but shouldn't it be treated differently from a "real" link. It's just a good example of where you could scream "nofollow, nofollow" in exactly the same way you do with paid links.
(5)Worse, it doesn't get at link networks that are developing where people are being approached to drop paid links within editorial content... Some people will get outed, but the search engines will never spot and discount all those links.
(6)The problem is, it's like propping up a house of cards. Now that so many are aware of the power of links, there's buying, selling, swapping, bartering, bombing -- you name it... That's one reason why personal search is so much being examined.

(1) yep, we’ve seen this recently, especially during and after Jagger once things “settled.”
(2) Once again, how can the SE’s determine intent, unless it’s the blatant red sports car you describe? Yes it is unfair.
(3) What if the ads are bought to pull link juice? You probably would have no problem replacing sponsors, but I can think of some sites/directories that would lose a chunk of change pretty quickly, which I guess is why this is such a lively debate.
(4) Right but if you don’t know about nofollow, wouldn’t a completely unrelated link be devalued anyway, in the “perfect algo?”
(5)MUhahahaha
(6) Yes perhaps personal search will become more used, and then we are all screwed. This would also give data to the SE’s that they can add to their algo to potentially give a boost to super-popular sites. How long will it take to create thousands or millions of “people” to all like particular sites?

[joeduck]

Danny - a superb summarizing post.

This debate gets at the heart of the key challenge facing most of us here and Google as well - how to rank properly in an ocean of sites designed to rank ... improperly.

Unless I'm reading Matt's comments wrong, I think Google is really trying to value something like the "sincerity" of the links - ie linking in an effort to help the user with no payments or other selfish intent. For this they probably look for older, long term, authority site links.

How should a perfect search engine weight the different types of links?

High value: Old non-profit authority site linking to old non-profit authority site in same sector.

Medium value: Reciprocal links between old established sites in same sector.

Negative value/ penalty?): Paid link to a porn scraper.

? value: paid relevant link. (Google is clearly telling us "zero value")
? value : paid irrelevant link

[PhilC]

Originally Posted by Chris Boggs

How on earth (or in the Googleplex bio/technosphere) could a search engine accurately (even 50% of the time) determine such intent? What if the link has equal SEO and traffic value?

They can't do it. I've written elsewhere that the best they can do is assume that links were paid for, and were bought for ranking purposes. Many or most of the assumptions would be correct, imo. For instance, if a site about seo places a sitewide to a hoodia site, the chances are the links were bought primarily for the rankings, so it's reasonble to assume it. But even if they were bought solely for the traffic, they would still be the sort of links that Google doesn't want to treat as genuine democratic votes.

Programmatically, they could spot plenty of bought links, or links they don't want to count, from the massive amount of links data that they have. It's not a tremendously difficult thing to programme.

[orion]

Happy to respond.

When I use the term "block" it conveys the notion of "segmentation", "portions" "sections", "compartments", etc; not to block or ignore anything. Since projectphp injected block level analysis into the discussion, I referred him to the two old SEWF thread that discuss this.

Block Analysis 101

Block Level Link Analysis



Nothing is ignored or "blocked". The whole idea of block analysis is to compute localized weights. True that in order to do that, anything that is not within the current block is "ignored", but this is done to isolate the current measure. Then you move to the next block previously "ignored".

The idea is not exactly new. It just happens that the MSN papers applied to the scoring of local links according to sections, portions, zones, areas of a document, you get the point (however, when you read the IR literature, the term "segmentation" and "passage" is given a technical meaning different from the semantic meaning we are adjusted to when we think of "blocks"). Block level analysis can be applied to any element, not necessarily to links.

If you check W3C DOM guidelines, they long ago mention how to structure markup so that specific elements are recognized as "blocks" of information. In particular they recommend the use of specific tags, (e.g, <map> tags and division/definition list tags) to add a localized block structure. In this sense, rendering a link menu as

<a>...</a> <a>...</a> <a>...</a> ... or as

<a>...</a><br>
<a>...</a><br>
<a>...</a><br>
.....


is not the same as rendering it using proper structural markup that exploits block levels and the HTML DOM structure . If you think that such placing of links is recognized as a "block", think again. If you check W3C they even tell you that. Does the expression "block elements" and "inline elements" rings a bell?


It is clear to me that many SEOs need to familiarize themselve with HTML DOM in order to exploit it to their benefit.

You think it is hypocritical to now assign relevancy weights, but then say we need a better model for scoring relevancy?

I believe I said that was in relation to assigning relevance weights to some links and not to others.

Incorporating linkage to the scoring of relevancy of written words simply compounded the problem of manipulating SERPs. Regarding an inheritance system, it is more of the same and will lead to more of the same. I don't buy it. Mr. Markety Spammy will only need to get married and form a "family" rather than buying real estate in a "link farm".

Yes, we need a better solution for scoring relevancy.

The way I see it now, the nofollow tag is like prescribing a band-aid to a problem that needs major surgery. While at least is better than sitting down and wait for the patient to die, still the band-aid does not prevent this. In the long run it accomplishes little and is more of a pc gesture (or I should say "becoming a pr nightmare"). The larger problem will still unresolved.

Orion

[projectphp]

Incorporating linkage to the scoring of relevancy of written words simply compounded the problem of manipulating SERPs

That I don't understand. AFAIK, there are only three types of ranking criteria:
1. On-page.
2. Offpage.
3. Human review.

If we remove one whole chunk of factors, the off-page, how would the results be better? Are there revolutionary breakthroughs in on-page ranking I am not aware of that could solve or substitute for off-page factors?

Short of detailed human review, and given people don't scale well, links are the "best" way to meassure all manner of "stuff". Without an alternative, links win by the two sweetest words in the English language: DE - FAULT.

The way I see it now, the nofollow tag is like prescribing a band-aid to a problem that needs major surgery.

So, just to be clear, is the issue is that nofollow is being used to attempt to solve the buying links problem, or that links themselves are the problem?

If it is the latter, see above, but if it is the former, I could not agree more! The nofollow link attribute was "designed", and I use that word in the loosest possible sense, to protect people that allow linking out by others, specifically blog comments. It was meant to indicate no editorial review, and that the link carried no weight as far as the sit owners were concerned.

That nofollow is now supposedly being touted as a way to indicate paid ads shows, IMHO, a problem with the whole darn idea. To summarize in advance: the nofollow link attribute is a terribly (un)documented and poorly thought out code addition, the use of which will lead to entirely unknown outcomes.

I have asked a Brazilian times, and no one can answer this for me: where is the nofollow unofficial documentation, and where is its official cousin?

As an example of actual documentation: robots.txt. this details how the implementation works, when and why.

Now for nofollow. http://www.google.com/search?q=nofollow shows a blog post first, and surely that isn't all Google offer! http://www.google.com/search?hl=en&l...og&btnG=Search indicates that http://www.google.com/webmasters/bot.html has some info and it does:

14. How do I tell Googlebot not to crawl a single outgoing link on a page?

Meta tags can exclude all outgoing links on a page, but you can also instruct Googlebot not to crawl individual links by adding rel="nofollow" to a hyperlink. When Google sees the attribute rel="nofollow" on hyperlinks, those links won't get any credit when we rank websites in our search results. For example a link,

<a href=http://www.example.com/>This is a great link!</a>

could be replaced with

<a href=http://www.example.com/ rel="nofollow">I can't vouch for this link</a>.

That is it! No "this is what the tag communicates", no "this tag means the page linked to will not be crawled unless found via an alternate link" and no "this tag indicates a paid link". It all focuses on the fact "links won't get any credit".

IMHO, the nofollow link attribute is an entirely undocumented code snippet with no stated or implied outcome or communicated information. The almost complete lack of documentation and expected outcomes makes nofollow worse than a de facto standard, as I have no idea, and no known way to find out, what the nofollow link attribute is supposed to communicate, nor under what circumstances it is to be used, nor whether the link will be followed,(will a page only ever linked to by nofollow be indexed?). That makes it a tag not worth a pickle in my book.

What is worse, comments like this by Matt Cutts make the situation worse:

What if a site wants to buy links purely for visitor click traffic, to build buzz, or to support another site? In that situation, I would use the rel=”nofollow” attribute. The nofollow tag allows a site to add a link that abstains from being an editorial vote. Using nofollow is a safe way to buy links, because it’s a machine-readable way to specify that a link doesn’t have to be counted as a vote by a search engine.

I would make the argument that "No Matt, that isn't what the code was meant to indicate", but really, what would I use to back up that assertion? A blog post? Comments from SES? IMHO, whatever Matt says nofollow does and why is right, because in the absence of any documentation, anything is true.

I think that nofollow needs to be both rethought and better documented, with explicit statements as to what will happen, what the tag communicates and suggestions on when to use it, to be documented somewhere, anywhere. If part of the goal of the nofollow link attribute is to indicate paid links, then so be it. Until then, I say it doesn't mean anything!

[PhilC]

To be fair, rel=nofollow does have one good use. It prevents blog/forum/guestbook spamming from being beneficial. It doesn't stop the spamming, but it does stop it from taking PageRank from the sites, so it can be helpful to websites in that respect.

It's been discussed in depth before, and, as far as I'm aware, no search engine has stepped up and said precisely how they (will) treat it, and it's not because questions haven't been asked - they have. Although it wouldn't really matter if an engine did answer all the questions. If all the engines don't treat it the same in every respect, then it isn't a standard or protocol, and it can't be treated as one by webmasters. Some people have serious reservations about the possible effects of it, and the engines need to get together and arrive at a standard that they will all stick to.

As it is now, it's the HTML oddity. Everything in HTML has a well-defined purpose, except the rel=nofollow attribute. Some attributes are used by some engines, and not by others, but they all have a purpose other than engines. rel=nofollow has no other purpose except engines, and it's effects are simply not known, because the engines seem to prefer to be tight-lipped about it. So be it. I don't use it for their benefit, and I won't use it for their benefit (that should bring them to their knees ).

I do think that Jeremy was unwise with his sold links, though. He's on the side of 'don't sell links' because it is detrimental to the serps. I agree with everything he's done as a webmaster, but he's not exactly a shining example of what his employer wants webmasters to do. His defense is fine for a webmaster, but I doubt that it's fine for Yahoo!.

[Gurtie]

I do think that Jeremy was unwise with his sold links, though. He's on the side of 'don't sell links' because it is detrimental to the serps. I agree with everything he's done as a webmaster, but he's not exactly a shining example of what his employer wants webmasters to do. His defense is fine for a webmaster, but I doubt that it's fine for Yahoo!.

I'm so glad he did. Perhaps now someone Matt Cutts and a variety of other Google people are aware of is doing something 'interesting' they'll have to start thinking like the rest of us....

If JZ's blog is now under suspicion and needs to be linked to with nofollow what about all those links people have been pointing there for ages? Will we, together with the Googleblog and YSearch blog, now be viewed with suspicion because we link without a nofollow? I mean I conducted due diligence before linking to Jeremy, I checked out his site and found it looked ok, in fact useful to my visitors. Now he's turned bad and I don't know what I'm meant to do.

It sucks...

So actually I happen to have heard about this and I can (if I can be bothered) go back and put condoms wherever I mention his site, but imagine how many webmasters out there find themselves linking to a site which later turns bad and have no way of knowing about it.

So really best if everyone in the world just adds nofollow to every link they make I reckon? That way no one will have to worry about it......

(Ok I confess, all I actually want is to see the wonderful irony when Ysearchblog has to nofollow their blogroll link to him and makes GHN not the only "we don't trust this site" link they have).

Nofollow, as a general rule, is stupid. If we link to sites its because we're happy to send traffic, link love, whatever else we send. If we don't want to then we just shouldn't link. Perhaps for commentspam or trackback spam nofollow is fine, but if we control the link then we should link properly and to be honest if I paid for a link and someone nofollowed it I would be a bit annoyed from a marketing viewpoint, put it through javascript or something no problem, but marking me as "a site not to be trusted" is not nice. I wouldn't buy an ad in a paper which added "*but we can't vouch for these people being honest" after the ad, would you?

The SE's either need to come up with a proper set of descriptive tags for different circumstances, and tell us exactly how they interpret those tags, or quit trying to change the way the www works.

[PhilC]

I wouldn't link to Jeremy's blog with or without a nofollow. He's a foul-mouthed person, and I wouldn't send people to read the stuff, but that's different.

[TheotherTim]

(Ok I confess, all I actually want is to see the wonderful irony when Ysearchblog has to nofollow their blogroll link to him and makes GHN not the only "we don't trust this site" link they have).

Hey Gurtie,
The no follow links pointing to GHN on the Y search blog were just a joke. We thought it was keeping with the spirit of the GHN site and were curious if anyone would notice. Anyway if you would like us to remove it just say the word. I enjoy GHN on a regular basis as you must see in the logs.
Tim
Ps. I will be communicating Yahoo Search's perspective on Jeremy's links in the near future.

[Gurtie]

lol - well it did make us laugh and gave us a good excuse to nofollow you back but you can make it up to us with an all expenses paid tour of the YahooPlex and by giving me that great opportunity to say "oi Matt, Yahoo take this nofollow stuff so seriously they use it to play jokes"

Ps. I will be communicating Yahoo Search's perspective on Jeremy's links in the near future.

looking forward to that....

[Daria_Goetsch]

Linking to other websites happened before Google arrived.

Paid banner ads were around before Google arrived.

The rel="nofollow" tag was created for a different purpose. It is now being used as a fix for algorithm problems.

[joeduck]

Tim - I sure hope Yahoo realizes what a great asset you have in Jeremy and his excellent tech blog, one of the most respected sources of technology info on the web.

I don't want to speak for Jeremy or Matt but it's obvious that part of Jeremy's motivation here was simply experimentation. This link debate has been very helpful in clarifying some things that Google clearly sees as "bad" even though the Google guidelines are conspicuously lacking in specificity on link issues.

I sure hope Jeremy doesn't get heat from the Yahoo side when he's helping the rest of us understand the big picture better.

[orion]

If we remove one whole chunk of factors, the off-page, how would the results be better? Are there revolutionary breakthroughs in on-page ranking I am not aware of that could solve or substitute for off-page factors?

Actually there are many current projects that perform as good or better without link-citation components when it comes to judging relevancy of words. Recent advances in non orthogonal term vector modeling (different from current vector models and good alternative to LSI) is one. Microsoft, Yahoo and even Google itself has other remedy projects under the hood to deal with the mess caused by incorporating linkage to relevance judgments. This often transpire at IR conferences. Yet, timing is everything here. Until then, we need to stay put.

Short of detailed human review, and given people don't scale well, links are the "best" way to meassure all manner of "stuff". Without an alternative, links win by the two sweetest words in the English language: DE - FAULT.

I challenge that. When it comes to sweetness, it depends who you ask. For some money is the default sweetest word.

I do challenge the perception that "links are the best way to measure all manner of stuff". It is time to let it go, link lovers.

And furthermore, I do challenge the perception that "a link is a vote of citation importance" similar to literature citation as many IRs from Stanford and opportunist marketers as well try to sell to the world.

As mentioned in the past in academic and marketing circles, anyone familiar with Dr Eugene Garfield's Citation Analysis knows that co-citation is driven by peer-review examination and editorial policies. On the Web, where anyone can say almost anything at any given time, or add, remove, update links at will or buy/sell links, link-spam a system, put/remove/activate/inactivate a nofollow attribute, etc peer review is absent, nada, zero, zip. Thus, it is fallacious to think in a link-literature citation analogy, at least for the Web "as is" now.

Links are coordination and jump mechanisms. Models based on these are great as scalable discovery tools and for crawling a web sub graph. But when it comes to make relevance judgments they are poor performers and introduce more noise than anything else to the measures.

A reader can make a relevance judgement of the piece that is before his/her eyes, whether the material satisfies his/her needs or provide a relevant answer to a question without no knowledge regarding how many link to the piece and why (for spurious/valid reasons).

Most of the time such relevance judgements are based on rationalizing the semantics of the current piece before our eyes, not its links. Isn't this what our experience as readers tells us?

If we perceive that the colored text placed in a link is worth a jump, we click and we jump to that piece. Then we start all over again and do a relevance judgment of the current piece. While the notion of passing semantic value between pieces via the jump mechanism appears to be actractive, is unpractical for sorting semantic value in a commercial environment full of noise and all sort of vested interest.

It is for this reason that better relevancy schemes are necessary. A happy medium seem more feasible than an appeal to pity.

I'm not afraid to try the following even if it fails:

Score and rank a Web sub graph with links and all, then remove weights passed through links (with or without nofollow attributes), recompute new relevance scores and re-rank results. If the sub graph remains relevant before and after, then node positions were well deserved. If not, recombine filtered nodes with a new sub graph. Iterate this process. This may minimize link manipulations and limit links to a discovery tool function. If it works, great; if it doesn't fine.

When you think thoroughly, this may as well be an experiment in the making.


Regarding Jeremy, that was a great experiment as well, as it can be seen from all player reactions. Funny to see others running for cover and saying "I will not touch your site without a condom because Google will do this or that...". Laughlable scene.


Orion