Keywords in urls

It is well known that Google highlights the words that have been searched for in the search engine result pages (SERPs) wherever they occur. This includes the URL and has led some people to erroneously believe that Google is therefore indexing keywords in urls.

Here are some examples of keyword highlighting in the url for the search |draw poker|. The first shows Google highlighting the keywords in a hyphenated url:

Keywords hyphenated in the url

This example shows the highlighting in a url with the keywords separated by underscores:

Keywords separated by underscores in the url

And this example shows the highlighting in the url where the keywords are conjoined:

Conjoined keywords in the url

Although Google is highlighting the keywords in the url this does not mean that Google is indexing keywords in the url. Let’s conduct an experiment to demonstrate this…

Three pages have been constructed with the url composed of three rather unusual words; callipygian, screak and quindecillion. They are all words in both Merriam-Webster’s online dictionary and Google’s lexicon. A Google search for them individually brings up the [definition] link.

A search for the three keywords together |callipygian screak quindecillion| produces less than a hundred results so when the pages are indexed for these keywords the pages will be easy to find. Each page has one link in and one link out and consists of three different sets of a 1,000 words of Lorem Ipsum so there will be no duplicate page problems.

The only occurrence of the keywords is in the url and you can see them here, hyphenated, underscored, and conjoined.

Will any of the pages be indexed for the search |callipygian screak quindecillion| and if so by which search engines?

I will write an addendum to this post in a few weeks when the pages have been indexed.

Addendum:

March 23, 2006

Google has now indexed the three test pages and the results are as follows.

For keyphrase searches:

Google test result 1

Summary: Hyphenated urls are indexed for hyphenated keyphrase searches and plain keyphrase searches. Underscored urls are indexed only for underscored keyphrase searches. Conjoined urls are indexed for conjoined keyphrase searches and hyphenated keyphrase searches.

If we search for a phrase that is unique to our three test pages we see the following:

Google test result 2

However if we search for the same unique phrase plus one of the keywords we see this:

Google test result 3

There is only one result which leads to the conclusion that Google does not index on keywords in underscored or conjoined urls but does index on keywords in hyphenated urls.

April 6, 2006

Yahoo has now indexed the three test pages and the results are as follows.

For keyphrase searches:

Yahoo test result 1

Summary: Hyphenated urls are indexed for hyphenated keyphrase searches and plain keyphrase searches. Underscored urls are indexed for underscored keyphrase and hyphenated keyphrase searches. Conjoined urls are indexed for conjoined keyphrase searches only.

If we search for a phrase that is unique to our three test pages we see the following:

Yahoo test result 2

However if we search for the same unique phrase plus one of the keywords we see this:

Yahoo test result 3

There are two results which leads to the conclusion that Yahoo does not index on keywords in conjoined urls but does index on keywords in hyphenated and underscored urls.

April 19, 2006

MSN has now indexed the three test pages and the results are as follows.

For keyphrase searches:

MSN test result 1

Summary: Hyphenated, underscored and conjoined urls are not indexed for hyphenated, underscored, conjoined or plain keyphrase searches.

If we search for a phrase that is unique to our three test pages we see the following:

MSN test result 2

Only two pages of the three appear to be indexed even if we click on “Show more results from…”. This is rather surprising because if we search on the site for a shorter unique phrase we can see that all three are in the index as follows:

MSN test result 3

I am not sure why this is or what could cause MSN search to behave in this way.

Even more inexplicable is a site search for a smaller unique phrase plus one of the keywords, depending on the keyword we see three different results!

For callipygian three results:

MSN test result 4

For screak no results:

MSN test result 5

For quindecillion one result:

MSN test result 6

These ambiguous results do not allow for a definite conclusion unlike Google and Yahoo. The best that can be said is that MSN indexes on some keywords in underscored, conjoined and hyphenated urls but that this is demonstrable only as a function of the search construction.

Status: Experiment completed.

Summary and additional information posted here URLs (Update).

12 Comments »

  1. protheus said,

    March 27, 2006 @ 4:04 pm

    Great experiment, i think it’s much better to try things out then to just speculate about this or that. Several years ago SEO was more of an experimental science because results were so much easier to get, you could modify a page in the morning (depending how good or linked the rest of your site was of course) and have it crawled, indexed and ranking in one day. Now results are harder to come by especially proving which of the many processes you did to get these results. I think experiments like these are important and i commend you for publishing your results. You need to come back to EE more often though you are missed over there.

  2. duz said,

    March 27, 2006 @ 7:59 pm

    >You need to come back to EE more often though you are missed over there.

    Thank you protheus for your kind remarks. How about instead of me going there, you send them over here :)

  3. Junanagoh said,

    April 7, 2006 @ 7:24 pm

    Great test! I think this has inspired me to try a few tests of my own that I have been waiting to try for a couple of months now.

    Ill be checking back to see the outcome of the test with MSN. Thanks for sharing your information.

  4. vingold said,

    July 27, 2006 @ 7:06 pm

    I’m a few months behind in my reading … but here’s my question:

    I understand that in regards to file names - hyphenation is the way to go.

    Can you speculate on what the results might be in regards to domain name?

    For instance - all other things being equal (exact in content, inbound and outbound links, etc.) - if I was to do a search on “seo blog”, which would rank higher?

    seo-blog.com or seoblog.com? I think we can assume you’d be less than efficient to do seo_blog.com

    I believe I read elsewhere in SEO literature (maybe even written by you) that in the end domain name plays a very small part in SERP. That all other things (meta tags, links, keyword file names, etc.) carry much more weight than the domain name.

    But I was just curious.

  5. duz said,

    July 27, 2006 @ 8:42 pm

    vingold that is a good question.

    Let’s get rid of the branding issue first.

    Ideally companies would like to build a brand and some of the most successful brands are made-up words, for example Exxon, Kodak, Epson, Comcast, Cisco etc., etc. If you want to build a brand that is the way to go and gobbledegook will do nicely.

    Although branding is very important, sometimes you may be more concerned about position in the SERPs. For example let’s say you are building an affiliate site for sports shoes. You have decided that your strategy is to have thousands of content pages of reviews of various sports shoes with your affiliate links. Branding is not an issue, so what domain should you choose sports-shoes-review.com or sportsshoesreview.com?

    Answer sports-shoes-review.com and here’s why. When your site gets unsolicited links they may be in two basic forms one where there is some keyword anchor text and one where the linker just anchors the url:

    For advice on choosing the right sports shoes go to <a href=”http://www.sports-shoes-review.com”>Sports Shoes Review</a>.

    For advice on choosing the right sports shoes go to <a href=”http://www.sports-shoes-review.com”>http://www.sports-shoes-review.com</a>.

    In the second case because you have used a keyword1-keyword2-keyword3 domain the anchor text will still contain your keywords. If you had chosen sportsshoesreview.com the search engines would not see the keywords but would see only one made-up word sportsshoesreview.

    So the bottom line is, more interested in SERPs than branding then go for the hyphens. Keep it to one or two hyphens only and make sure it reads well as the name of your site.

    Will it make a difference? Yes but a small difference. Is it worth doing? In most cases yes and I would take http://www.sports-shoes-review.com/nike-shox.html over http://www.sportsshoesreview.com/pdp.jsp?productId=99057&categoryId=304203&productGroup=90793&catalogId=1 every time.

    Sports Shoes Review sounds good as a site name doesn’t it and sports-shoes-review.com is still available as I write :)

  6. kirtan said,

    September 13, 2006 @ 2:25 pm

    duz,

    I have the same question as vingold has, but slightly different one.
    Say one site is with its brand name (e.g. www.nike.com) and one with keyword (e.g. www.sports-shoes.com) And everything is same (meta tags, title, link pop, etc) which one is get top rank for the keyword “sports shoes”. ?

    i guess www.sports-shoes.com will be top rank than branded one. How much effort needed to out rank that unbranded site?

  7. duz said,

    September 13, 2006 @ 4:33 pm

    That is a very hypothetical question but if everything was the same then sports-shoes.com would probably rank higher. Of course everything is never the same in the real world and brand names are extremely important to businesses. The top site for online books is amazon.com not books-online.com, the top auction site is eBay.com not online-auctions.com and so on with Microsoft, Coca-Cola, Disney, Cisco and thousands of others. If you are building a business then you need a brand name. If branding is not an issue then go for the hyphens.

    I think in practise the same effort has to be put into emarketing your new website whatever domain name you choose.

  8. Kirtan said,

    October 5, 2006 @ 1:21 pm

    My site http://www.gx-affordable-web-design-and-hosting-services.com has #1 rank in MSN and Lycos now for this keyword phrase “Affordable Web Design And Hosting Services”. But there is no rank in yahoo and google upto 1000 position. Is it because of LONG URL. Or any other reasons exists for this?

  9. duz said,

    October 5, 2006 @ 4:17 pm

    It is an over hyphenated domain name but that is not the reason I suspect. Normally two hyphens in a domain is the most you should have and something like genex-web-design.com would have been better.

    It looks as if your site maybe sandboxed. For example doing an exact search in Google for a short text phrase taken from your home page |”located in new york and india”| does not find your site. The way out is to obtain some links from “trusted” directories like dmoz and trusted sites in your sector like webdesign.about.com.

  10. Kirtan said,

    October 5, 2006 @ 4:35 pm

    Just today i updated my site with W3C xHTML validation and changed content. Google locate my site for following phrases | “still it suits to your budget” | and |”product familiar to your”|. So i guess it is not sandboxed problem. is it so? Or i have to get some more quality inlinks? so far i didn’t spend any money in buying links. Because it is not upto 1000 Ranking and how can i guess, it will come up if i buy links? confusing. Please Help! Thank you in advance.

  11. duz said,

    October 5, 2006 @ 4:57 pm

    I think it probably is sandboxed or very close to it and you will need some “trusted” links. You are right not to buy links at this stage. Just work away at getting those links and be patient.

  12. John Scott said,

    November 25, 2006 @ 8:54 am

    Coincidentally, I blogged about this just a couple days before you:

    http://www.internet-marketing-blog.com/2006/03/10/parsing-keywords-in-urls/

    Great minds think alike, eh? ;)

RSS feed for comments on this post · TrackBack URI

Leave a Comment

Bot-Check