Update 2008-07-09 It’s become clear from the comments on Sphinn that people are not paying attention to what I actually wrote below. I did warn you that the search results might change for a variety of reasons, AND I said I would take the links down. I also said that this was not a valid test. All I did was show that the other tests are not valid.
Update 2008-07-10: For those of you in the SEO community who don’t know what a “flame” is, here is a time-honored definitition of what is a flame. It more-or-less says that comments like, “So-and-so is wrong 9 times out of 10,” are flames….
2nd Update 2008-07-10: For those of you who are curious where Google stands on the subject of second links passing anchor text, Matt Cutts has said “Typically if the anchortext on the two links is identical, we would probably drop one of those links.”
So Rand Fishkin and Michael VanDeMar have launched and perpetuated another SEO myth: that Google will ignore the anchor text from a second link on the same page. That is, if you place a link with anchor text A and a link with anchor text B, both pointing to the same destination, on the same page, Google supposedly only passes the anchor text from the first link.
When David Eaves showed that Google.UK will pass second link anchor text, SEO Scientist argued this is not true for Google.com. SEO Scientist’s blog post was more convincing than the arguments made by Rand and Michael, but his test was still flawed.
In the comments to that post I suggested that — as described — the test was not valid. To construct a valid test you need to create conditions that anyone can replicate, and that includes the type of Web sites you’re using. People have not been disclosing the sites where they place the links so it’s impossible to determine what sort of conditions are really applying to these tests.
A quick analysis test
On Friday, July 4 — as a lark — I dropped two links to SEO Theory onto the front page of Xenite.Org. Xenite is a trusted domain that has been around since 1997. SEO Theory has earned a fair amount of trust in Google thanks to all the links you good people have pointed to it over the past year. I’m reasonably certain that the home page for Xenite (http://www.xenite.org/) and the main blog page for SEO Theory (http://www.seo-theory.com/wordpress/) are recrawled on a daily basis and are usually recached once or twice a week.
Frequent recaching is a good sign that a page is included in the Main Web Index rather than the Supplemental Index, but as I don’t work for Google I cannot provide you with a definitive explanation of how they implement their dual index strategy. Nonetheless, if you have two sites with similar crawling/caching profiles you should be able to perform this test for yourself, but you definitely need to know how often your sites are recached.
Initial Test Results
By Saturday evening (July 5), Google was showing both pages in both queries. That is, at least temporarily, Google has been passing anchor text from two links on the same page to the same destination. However, Google’s weekend search results tend to be a bit unstable and things usually change by Monday as Google integrates new data into its index. I cannot guarantee we’ll see these results continually (or that they appear in all data centers). So I’m providing screen captures here to document what I have seen (keep in mind that I don’t particularly like screen captures as they can be faked, but I swear I have not faked these).
I have to use screen captures here to show you the queries as trying to embed them on this page without making it relevant to those keywords would be a real challenge. Note that I used hyphens to connect the words in the queries, rather than embedding them in quotes. The hyphenated query expression works a little differently from the quoted Exact Find mode query expression. I use hyphenated queries to find recently crawled but not yet fully indexed content. I would rather not have shared that secret but Neyne (SEO Scientist) was puzzled by the query links I posted in a comment to his blog.
Screen capture 1 shows you the first query expression:
Screen capture 2 shows you the second query expression:
Screen capture 3 shows you the search results for the first query expression:
Screen capture 4 shows you the search results for the second query expression:
Is this a valid test?
As far as answering the question, “Does Google pass only the first link anchor text” or “Does Google pass anchor text from a second link”, this test is not really valid. First of all, I used only one set of links on Xenite. I didn’t have much time to do anything and, frankly, was almost convinced by the SEO Scientist blog post that maybe Google really doesn’t pass second link anchor text (which is a bizarre idea but as an experienced programmer I can think of plausible reasons for why that might be so).
So there are no control links to compare this set of links to. Also, the timeframe is very, very brief. The search results may not appear the same over the next few days. In fact, I will remove those links from Xenite’s home page before the end of the week, so this is a very transient test. A much better test would keep the links in place for at least a few months.
Also, I haven’t gone through the reverse-and-redo steps that I recommend people do with their tests. Hence, this test is not a valid cause-and-effect test.
Nonetheless, as a quick “Let’s see what happens” sort of experiment, it very quickly demonstrated that, yes, Virginia, you can get Google to give credit for a second link from a page to the same destination — at least temporarily if not permanently. It really depends on the quality of the source and destination pages, in my opinion.
The two expressions do not occur anywhere on SEO Theory, although the individual words do. Nonetheless, you can use the intext query operator combined with a site query operator to see that the expressions have not been indexed on this site. I cannot promise that I won’t use them on this blog in the future, so you may eventually find them through site queries.
Debunking Rand Fishkin on intext
At last month’s “Give It Up” session in SMX Advanced Rand Fishkin reportedly said the intext query operator is broken, or is crap, or something (I read a couple of blog posts about the session on July 3). Here’s a little clarificaton for Rand and everyone else: The intext query operator only works with indexed content.
That means that if you’re looking for specific text on a page that is in Google’s Supplemental Index, you’re not very likely to find that text using the intext query operator — although Google does index SOME text from Supplemental pages (particularly rare words), so you may get lucky.
A possible test to show you which pages are Supplemental
Which leads us to the conclusion that you may still have a valuable means of determining which of your pages are probably in the Main Web Index: use a site query combined with intext to search for expressions you know are on your page (unique expressions work best). Then look at the cache date for the returned page(s). And keep your expressions to 5 words or fewer.
If the cache date is pretty old (more than 30 days) you’re not looking at a frequently cached page. If you don’t get any pages in your results at all, that’s a pretty good sign they are in the Supplemental Index (or just not indexed at all).
You should use the hyphenated query expressions carefully. To be honest, I don’t like revealing these kinds of tests because if the SEO industry talks about them ad nauseum Google may disable the query operators (I’d rather Google just be honest again and restore the Supplemental Results label, or else give us a query operator that shows us what the Supplemental Index pages are for a site).
Final word about Michael VanDeMar
Mr. VanDeMar has been bashing me on various occasions for at least a year. He made the comment on the SEO Scientist blog that you can safely ignore me because 9 times out of 10 I’m wrong. Frankly, after being flamed and attacked by him numerous times on SEO Refugee I just stopped trying to respond to his nonsense.
People who resort to unprofessional behavior and poison pen campaigns in order to win arguments will never impress me as being knowledgable or trustworthy. Anyone who wants to believe Michael VanDeMar’s assertions about me, PageRank, or anything else is more than welcome to. But I will certainly appreciate people ignoring all the flame-bait and insults and reaching their own conclusions (about how often I may be right or wrong) based on the evidence I provide.
There are a lot of people in this industry who are emotionally married to specific ideas and SEO bloggers. They won’t change their minds no matter how often you debunk their myths. You cannot help such people. All you can do is remind yourself that inflexibility on any point makes the other guy less likely to be a competitive SEO, which always works in your favor.
See you in the search results.
{ 12 comments… read them below or add one }
sjachille 07.06.08 at 10:41 am
That was a good one – yet again I think the main outcome from you very nice post is that techniques come and go over the years, but good down-to-earth SEO in a broad and not-so-technical implementation are probably teh best way to go about optimizing a website.
In fact I have sites ranking for years and I myself rank very well for pretty competitive keywords over on this end o the pond (in case you are wondering I am Italian but with an international background).
And yes you are right people get too involved emotionally and are side tracked by the noise online: More often than not many are more simply seeking ways to get top rankings without any sweat, then cashing in but it doesn’t work like that – at least not in my experience and I have been around in this business right from the start in 1994.
seo_scientist 07.06.08 at 12:13 pm
Michael,
I am still not convinced by your test. Your testing setup is not even close to being sterile. You have pages on your site that include every one of the tested keywords. They still come up for me in the SERPs, especially the page where you are linking to Matt’s SEO Documentation post. All of those pages are linking to your homepage. Your test is dirty.
As for your supplemental index claim. Both of the sites I have tested the theory on are being indexed frequently. The largest span I had between the indexes is about 10 days. Using your site: intext: combination, I get June 30th on one and July 3rd on the other. So, according to your test, these are not supplemental. They are both homepages of domains that rank on the first or second page for some mildly competitive phrases. Their inner pages rank as well.
I have created a testing environment, observed the effect, reversed the conditions, reversed the effect and repeated the process with consistent results. The test is now entering its third month and there is no change in results. The second link is simply not passing value.
And no, I will not disclose the testing sites. Mainly because I have pages there I use for linking. I am not disclosing my linking network.
It was not my intention but I did the test according to all of the limitations you have outlined here over several posts. And still, according to you, my results are not valid, and your support for that claim is your contaminated test. Had someone else setup a test in a conditions like you did, you would have laughed at them. Now who is the one married to an idea ? And Michael’s obsession with you ? How many posts addressed at Rand do you have ? Do you see the pattern here ?
Anyways, I am continuing to read your blog, because you have good posts and great line of independent thinking. It takes a while to filter out the condescending and the personal remarks but the rest is OK.
underworld 07.06.08 at 12:30 pm
google definitely uses as much information as it can get, second, third, forth link text etc.
Michael Martinez 07.06.08 at 4:46 pm
seo_scientist: “… You have pages on your site that include every one of the tested keywords….”
Michael: There are, as I pointed out in the post, reasons why the test is not sound but your concern is not one of them. So, no, the test is not “dirty”. It’s merely incomplete but in its incompleteness it nonetheless shows that Google can and will pass anchor text from a second link — sometimes. Sometimes it won’t. Explaining why the effect is not consistent requires more work than you and I have brought to the subject combined.
As for the specific phrases I used, the hyphenated query only returns exact-phrase matches and the phrases do not occur anywhere on this blog (I searched it to make sure). Finding blog pages for individual words does not invalidate the test, but you have to understand how that query expression works to see why. It’s not the kind of query expression people in the SEO community are used to.
seo_scientist: “As for your supplemental index claim. Both of the sites I have tested the theory on are being indexed frequently….”
Michael: I cannot comment on the status or quality of the sites you refer to, as you have not disclosed them. I do appreciate fully the need to respect client non-disclosure; nonetheless, without disclosure we cannot trust any of the claims that people make about how search engines behave. I have used two sites that are very public and very well known. Whether I spend the next 3-6 months dropping and re-adding links is irrelevant. Anyone who is curious can look at the sites and the search results for now.
I remain skeptical of the assertion that Google only credits the first link. There are, as I have explained in the past, many possible reasons for why various SEO linking tests fail. In that post from last month I wrote:
My incomplete test not only uses two indexed trusted pages, it uses two frequently indexed trusted pages — a condition I did not specify.
However, I also wrote:
I am very familiar with the pedigrees of both Xenite.Org and SEO Theory. I built the sites from scratch and have used them both to test many SEO hypotheses. I have a pretty good idea of what I can expect to accomplish with them. These are not client sites, but rather sites over which I have full and complete control.
I don’t have to wait a month to see if these pages are stable. I already know they are.
What I did this weekend doesn’t prove anything so much as it merely disproves the assertion that three people have gone on record as making: that Google won’t pass anchor text from a second link. That simply isn’t true. What may be true, and perhaps may be worth investigating, is that Google might eventually drop the second link’s anchor text. Or maybe not. If I had more time for this particular topic I could devise a long list of tests.
Frankly, I cannot imagine many reasons for why someone would want to pass anchor text from two or more links on the same page, unless they were trying to be stealthy. It might fool some of your competitors if you break your anchor text up but some people will dig deep enough to find what they are firmly convinced is there. If they are right, they’ll find it.
seo_scientist: “… Had someone else setup a test in a conditions like you did, you would have laughed at them. Now who is the one married to an idea ? And Michael’s obsession with you ? How many posts addressed at Rand do you have ? Do you see the pattern here ?”
Michael: I would not laugh at anyone who disclosed their linking source and destination while everything could still be inspected by third parties. That’s something you just don’t see the SEO community do very often. As for my posts rebutting Rand Fishkin, it’ s true that I do occasionally debunk things he writes about. The difference between me and Mr. VanDeMar, however, is that I back up my objections to Rand’s comments with substantial explanations and I don’t just say things like, “Rand Fishkin is wrong 9 times out of 10 — you can safely ignore whatever he says”.
There’s a pattern in Mr. VanDeMar’s attacks on my character that you won’t find in any SEO Theory blog posts, except for the numerous times I’ve vaguely referred to various “SEO idiots” out there in the blogging community.
I leave it for the reader to decide who best fits that description without insisting they match my own personal definition.
In the meantime, I stand by what I wrote above. Google clearly passed some anchor text from a second link, as least in a transient way. I make no promises about whether it will give full credit for any length of time.
But this subject is, for me, closed.
randfish 07.07.08 at 9:31 am
Michael – just FYI, you can disagree and call me out whenever you like. I think the butting of heads on opinions and on search engine operations is excellent for the search marketing community and furthers the goals we all have to be better at our jobs. I certainly appreciate it when you’re gentle with the criticism more than when you’re harsh (don’t we all), but regardless, I wouldn’t want you to stop thinking critically or posting your thoughts.
kinetic 07.07.08 at 10:40 am
Michael VanDeMar is supposed to be a big name in the SEO industry?
Sometimes, the only way to be at the top is to hit on the ones already there. If He is doing this for a year now, i think it’s maybe time for him to change is the game plan because it’s not working…he is still very very unknown..sorry to say the truth but this guy is not credible…so why pay attention?
SEOHolicc 07.09.08 at 8:38 am
Great post. I just had to comment on your very last paragraph though. That describes so many people in this industry to a T, and I thought it was very well put.
Local SEO Guide 07.09.08 at 9:07 am
Michael – I for one appreciate the effort to try to figure this stuff out. Every little test adds to the base of knowledge out there.
Michael Martinez 07.09.08 at 10:00 am
I see Neyne has claimed some sort of victory on the Sphinn discussion of this post, so apparently all objectivity has been thrown out the window as egos collide. There is nothing Popperian about the call for disclosure.
This topic probably won’t fall into the “Settled Disputes” category for a long time.
seo_scientist 07.10.08 at 7:25 am
Michael I really don’t have time nor strength to get into power struggles with you. No one is claiming any victory, this is not my personal struggle. You think you have debunked the theory, I claim you didn’t. You are obviously not accustomed to people disagreeing with you and that is really not my problem. In the end, everyone will believe what they find most reasonable. I am fine with people thinking you are right and I hope (for your own peace of mind) that you are fine with people thinking you are wrong.
Take care
Branko
Michael Martinez 07.10.08 at 7:32 am
Neyne, I’m not looking for power struggles. Nor am I claiming to have debunked a theory. I’ve merely shown that the tests performed so far are not valid. Your personal comments are both unwelcome and unnecessary.
One does not graciously disagree with another person by telling people that person cannot handle disagreement.
I still don’t see much value in the question of whether a second link passes anchor text. At best, we have collectively shown that while it can happen most people don’t see it happening. I think the SEO lesson to be taken away from all this is that people should not be counting on two or more links per destination on a page.
seo_scientist 07.10.08 at 7:43 am
Great!
You must log in to post a comment.