Hi SEOmoz folks,
Some weeks ago my coworker Leandro Riolino published in our blog an experiment he was working with. The idea of the experiment was to try link to a page A from a page B with 3 different anchor texts providing value of all those anchor texts.
The idea is simple: we chose 3 random keywords, created an internal page, created 3 links to different URLs that have a canonical tag to the main page. You can see this idea illustrated bellow:
So, after choosing the 3 keywords we submitted each one to check if Google has any occurrences of them:
Then we bought a new domain, that has no backlinks and as you can see bellow, Google shows us that this website isn't in the index:
Creating the Index Page
To start the experiment my coworker downloaded a random template from the Internet with some random content inside, changing only the page title, meta description and H1 tag focusing all them into the main website keyword “jogos online de corrida” (online race games in English). The major change he made into the template was to add a conditional check with PHP to insert the canonical tag if the URL requested had any parameter:
<? if (isset($_GET[keyword]))
{ ?>
<link rel="canonical" href="https://www.jogosonlinedecorrida.com.br" />
<? } ?>
For those who know something about PHP language, this code checks if the variable $_GET exists. If this check returns true the code insert the canonical tag line into the HTML.
It’s important to say that we do not mention any of those 3 keywords in the Index Page. So, this page can’t rank for having a keyword mention… instead Google needs to check it’s backlinks.
Internal Page
The next step was to create the internal page. We created it with 3 links in 3 different page positions: one in the header, another one in the content area and the last one in the footer area with the following anchor text: “nanuoretfcvds ksabara1″, “esjstisfdfkf aasjdkwer” e “gisrterssia fdswreasfs”. Each link had different targets:
- https://www.jogosonlinedecorrida.com.br/?keyword=key1
- https://www.jogosonlinedecorrida.com.br/?keyword=key2
- https://www.jogosonlinedecorrida.com.br/?keyword=key3
It’s important to say that we used the meta tag <meta name=”robots” content=”noindex,follow” /> into this internal page, so this page would not rank for those 3 keywords.
Indexing the Content
In order to have the pages indexed by Google my coworker created a Sitemap.XML with the 2 pages (home and internal) and submitted it to Google Webmaster Tools. It is important to say that we did not share this page in any webpage and did not submit in any bookmarking service.
After 2 weeks, our website was showing the 2 pages when we used the operator “site:”. After one more week Google was showing the 2 pages and the link to their cache.
After this “waiting time” we searched in Google on the 3 keywords that we created and noticed that the main page was appearing for ALL of them as you can see bellow:
So, with this small experiment we noticed that Google was giving to a page 3 anchor text values if we use the canonical tag as a funnel.
Conclusions and Applications
With this small experiment we have a hint on how Google treats the anchor text of a page that uses the rel=canonical tag and now we can try to create some new experiments (eg.: use a parameter in the logo link to your main page, and then receive the anchor text of the second link – because we know that only the first anchor text counts).
We know that this is a single experiment and we need to see if this works in a real website, because we know that Google understands the page segments and this maybe does not work as we presented in this article. We still need to try and check this.
I can’t end this article until saying congratulations to my coworker Leandro that provided me a huge amount of knowledge with this experiment – thank you.
Hope you liked this article!
I ran the same experiment over a year ago. It worked then as it works now. See Getting more links from high ranking pages
great blog James...nice one
Great post and great intuition. I agree with you that this first experiment has to be followed by others, maybe using (safely) real websites.
On a more lighter hand, I had to thumb you up also because of the domain name choosen... in Spanish it sound so like "Online Games about Bullfight" ;)
Thanks Gianluca!
Btw, we own lots of rich domains :P
An interesting idea and well thought out approach to testing your hypothesis. However, I'm really not sure that this tactic is entirely beneficial. I am really interested to hear your thoughts on a couple questions I have.
Here are my questions:
A) Yes, you now have three different anchor text links to the home page, but aren't each of those links now diluted because the internal page has three links to the home instead of one?
B) Might the use of canonical tags further eat a tiny bit of link juice? In other words, would a single direct link from the internal page to home carry more link juice than the sum of all three canonical links?
C) Finally, I'm not entirely sure how the connection between anchor text and link juice works in detail. Say I want to rank well for "blue widgets". Using your technique, I set-up three links from an internal page with the following anchor text: "blue widgets", "red widgets", and "green widgets". I suspect that I would be better off with a single link from the internal page with the anchor text "blue widgets". Am I right?
Thank you for an excellent, thought provoking post.
Great article, great questions.. looking forward to the answers :)
Actually, as I said in the article, we don't have any answers to deep questions. We just found this weird.
What we are trying to check w/ new experiments are your questions A and B.
When we talk about giving more than 1 anchor text to a page, you should consider that sometimes we do not target only 1 keyword in a page (like a homepage). Sometimes you will have a link in your logo and than, in your content, you need to link to your homepage. How can you give some value with that? This is what we found to try.
As soon as we find answers to your questions we will post here.
Thanks for your comments.
I fully agree Fabio, this is a great tool to have in the toolbox for certain situations. Kudos to to you and others for publishing this method and your test results.
Amazing post! I love the experiment model, and you've done it brilliantly. I wonder how this will work for competitive keywords or possibly even misspelled keywords?
Looking forward to the next post.
About mispelled words, isn't Google getting better in armonizing the search results for those kind of words?
I think it works really well. The point is that when someone search for a misspelled word, Google will correct the results showing the non-misspelled results page. I think using misspelled will not help you but if you use considering plural vs. singular could be helpful.
This would be an incredibly valuable use of this in my eyes. Thank you very much for sharing this. Really smart stuff!
Hello
absolutely right one. if you r using wrong keywords it will not helpful to you but if you r using another word having same meaning definitely u will get a good response.
Many thanks for this amazing information...again..thanks!
Fabio! This is pretty clever. I'm impressed. You're the man!
Sometimes a page can only rank for so many keywords before it starts looking like a stuffed pig. This might be an interesting technique to use to beef up your long tail targeting. I'll try it!
Pig and beef mentioned in a single comment. Top that!
Must agree, very interesting stuff!
This is certainly useful research and could make it easier to rank for some longtail keywords. Although it might look a bit spammy when you overdue it- like most other things. I was hoping to publish something similar (different approach though) hopefully soon, the overwhelming feedback to fabio encourages me even more. Thanks!
Interesting post, but I'm a little skeptical of the benefits vs. costs. Repeated instances of this linking strategy could be seen as an abuse of the canonical attribute and could eventually hurt rankings because it's meant to reduce multiple content issues, not perpetuate them. Maybe I'm getting this all wrong, but doesn't this tell Google that it needs to hone in on all 3 links that include parameters AND the same content? It forces the bots to index 3 versions of the same page, when the proper use of a canonical attribute would be to tell it to use 1.
Could you do the same thing with a 'title="keyword"' attribute and not force multiple versions get indexed? If I have the details wrong here, please disregard.
I agree. It seems like a spammy strategy to say the least.
It also demonstrates to Google that your site structure needs cleaning up.
There are spammy and non-spammy applications of this. The primary point of this is to realize that as of right now, anchor text seems to be passed through canonicals. It's a big deal.
As soxhead said above. You have spammy and non-spammy ways to use this idea. We found that this ideia works, not that you should use it everytime.
And we did not conduct any tests with title attribute. I'll try to do that in the next days. Thanks for all comments.
Excellent experiment and discovery, Fabio, thank you for sharing over here with everyone else. ;)
Great post!
But it would be nice to have a control group where you linked directly to the target page like that to see if maybe Google would rank the target page without these canonical pages.
Do you mean try to compare which version Google likes more?
I think he means just link to the canonical URL without the keyword args --- what's the difference between this approach and just linking to the URL using different anchor texts?
Or, what happens when you put the anchor text into the keyword argument, so it becomes part of the URL?
A similar trick works when you put keywords into the canonical URL you want to associate with the page, BTW.
For example, if you have the page:
https://domain.com/h76qga.php (random URL)
which is also mapped via .htaccess as
domain.com/aaaaa/
and the canonical tag is domain.com/aaaaa/ ... the page will be associated with 'aaaaa' via URL even if no anchor text points to it.
Etsy does this in product-page URLs.
The arguments are used to create a new URL, so the canonical make sense.
If you put keywords in the URL you are not measuring the canonical tag, but you are giving another way to Google identify what that page is about. Got the idea?
Absolute rockstar! We should all do more experimentation and less guessing.
I was wondering if someone could tell me why this code segment was added to the homepage:
<? if (isset($_GET[keyword]))
{ ?>
<link href="https://www.jogosonlinedecorrida.com.br" />
<? } ?>
What is the difference between having the canonical ONLY on the GET parameter version of the page as opposed to having it on the normal https://www.jogosonlinedecorrida.com.br page as well?
You can use the canonical tag point the correct version to itself. But for a really "clean" code we prefer to not add the canonical in the correct version.
Wait a second. There is no gain. It's a loss.
If the page I'm linking from has PR5 and I have only one link (anchor1) I get all the juice for that anchor1 (maximum possible pr). When you create more than one link and "cheat" the search engines you divide the juice among multiple anchors. Your maximum possible pr is the same (pr5). No improvement. I would argue you made your backlinks worse as you lost the juice for the anchors.
Hey Fábio nice post man, this is a great idea, if u would like to try in my website we can, I have about 30000visits/day.
thx to share with us.
Thank you so much for that experiment. It gives us a better understanding about how Google manages inbound links, anchor texts and canonical link elements. I will bookmark this post !
Its really great idea. now i can give 2 or 3 ancher text with canonical tag. thank you for this technique.
Thanks for the post, really good!!
Really great post. You've shown how a simple experiment can be run on a small site without the need for a lot of number crunching or 1000s of test pages. Classic SEO at its finest!
The important takeaway from this is not that you can manipulate anchor text distribution, but that the canonical tag passes on whatever anchor text is pointed at it. Yet another smart way to take advantage of the canonical tag.
I like this experiments, they are so useful in predicting the search engine trends. My question is: Why did you guys take a long time like 2 weeks to wait for the results. It should be quicker that that right or did you have any other reason for that?
H
We waited because that was the time Google took to find our experiment.
Great post! I find experiments like this so fascinating. Looking forward to seeing the followup and trying this out for myself.
- Evan
So if I understand you correctly, what you're saying is, If page A has three identical links that link to a page, only the first appearance of those 3 links will pass anchor text value?
Therefore, the way to get all 3 links to pass anchor text value, is to make all three links unique? and one easy way to do this is to append tracking strings at the end of the links.
So, instead of
Link 1 = www[dot]domain[dot]com
Link 2= www[dot]domain[dot]com
Link 3= www[dot]domain[dot]com
Do something like
Link1 = www[dot]domain[dot]com/?utm_source=citysearch&utm_medium=banner&utm_term=jabberwocky&utm_content=hamsters&utm_campaign=nuisance
Link2 www[dot]domain[dot]com/?utm_source=citysearch&utm_medium=banner&utm_term=jabberwocky2&utm_content=hamsters2&utm_campaign=nuisance2
Link 3 = www[dot]domain[dot]com/?utm_source=citysearch&utm_medium=banner&utm_term=jabberwocky3&utm_content=hamsters3&utm_campaign=nuisance3
Each link could have the same (or a different) keyword as anchor text, and each link would transfer the anchor text value to the linked-to page?
Am I understanding this correctly?
Yes, you got it, although it doesn't have to be tracking code that is appended.
Thanks, dvansant, for confirming. Yeah, I just used the Google URL builder just for speed sake. Just making sure I understood what was going on. That's why I love SEOmoz: the info makes me think (hard, sometimes) lol!
This is hot! I see some major link architecture restructuring happening for our webmaster this weekend!
amazing experiment.
Actually this is not something Quite new. I publish the test something like a month ago ;)
https://bit.ly/9TQqZR
I'm a little confused by this. Are you only making additional pages in order to canonically direct a crawler to the homepage? Why not just make 301 redirects, since canoncial tags work in a similar way? For example, you could make domain.com/keyword1 redirect to the index page and build anchor text links for keyword1 pointing to the /keyword1 page. Wouldn't that be a more elegant way of doing the same thing?
I'm afraid you have misunderstood what Fabio has done. Try reading Petrosianii's summary explanation a couple comments up. He provides a good summary.
I could be wrong here, but if I remember correctly Google would index your "3 pages" even without the canonical tag. As long as your link is using a different url they will be valid links on the same page. In other words your query parameters make the backlinks different and therefor your can tag was completely useless.
I always create two backlinks from one page like this:
https://www.domain.com/
https://www.domain.com/default.html
Google will pick and count both links for PR.
In this experiment, we only expected that Google will index the canonical, not the "3 pages". Perhaps in your example Google already used https://www.domain.com/ as the canonical, even if no canonical link element was used. If this is what you meant, then Yes, sometimes there is no need to add a canonical link element because Google picks the canonical itself.
I liked how you included screen shots. Thanks for sharing!
Page rankn/a Linking to other blogs can be a real time consuming project. There really are no shortcuts to this process but with hard work and effort, you can have a "Domino Effect" of comments on your blog if done correctly.
Thank You,
D Lee
[link removed]