A Good Alt is Better Than a Bad Anchor

11 Mar

A short while ago, I set up a small test in order to determine the value of an alt attribute, in comparison with a title tag and anchor text. Although quite a few link building text books advize you to avoid using image links, and the results for this subject in the Link Value Factors were inconclusive, I have always believed that optimized image links can be nearly as valuable as regular text links.

Tests in the past, as well as experience, have showed us that if you compare a text link with an optimized anchor text with an image link with the same words attached as alt attribute, the page where the anchor text link points to will rank higher. This means that there’s no need to determine if a text link is “better” than an image link, but it still is worth to test the differences and smaller factors underneath. Nothing big, but useful enough to run a test.

The test

For this test, I have set up a page on a new subdomain of an existing domain. Unfortunately, I can’t share the domain, nor the exact keyword combination, because this particular test has already been (accidentally) terminated. Let’s just say I deleted a file I wasn’t supposed to :)

First, I wrote 6 different texts that were optimized for a combination of 2 different keywords ([KW1 KW2]). Both of these keywords return results in Google, but if you add them together in a single query, zero results show up. All texts were unique, but similar in layout and optimization technique. The keywords where the pages have been optimized for were in the exact same locations (word-wise) on every page.

Then, I set up a page with 6 different links. Three of these links were image links, and three were text links. For the images, which all had non-keyword filenames, I used different images of the same size. Although I do know that Google uses link order in the process of link valuation, I left this factor out on purpose.

The positions of the links on the page were:

  1. An image link without an alt attribute or title tag.
  2. An image link without an alt attribute, but with a [KW1 KW2] title tag.
  3. An image link with a [KW1 KW2] alt attribute, but without a title tag.
  4. A text link with a non-keyword (ie click here) anchor text.
  5. A text link with a partially optimized (more about KW1 here) anchor text.
  6. A nofollowed link with [KW1 KW2] anchor text.

After all pages had been uploaded, I added two links for indexation and let the test soak for about four months. During these months, all results were constantly (every week) monitored to see if the rankings were stable. Just to be sure, I also set up a control test, but that one was slightly different from this test.

The hypothesis

Like I mentioned, I do know that the order of the links matters. Although the differences are small, it does matter if you’re the first link on a page or the 217th. Because it is extremely difficult to set up a page in a way that it distributes an equal amount of link strength to each link, I decided to list the links in an order that would make any positive result clearly visible. The -in my opinion- least valuable link was listed first. I expected Google to list the pages in the following order after searching for [KW1 KW2]:

  1. An image link with a [KW1 KW2] alt attribute, but without a title tag.
  2. A text link with a partially optimized (more about KW1 here) anchor text.
  3. An image link without an alt attribute, but with a [KW1 KW2] title tag.
  4. An image link without an alt attribute or title tag.
  5. A text link with a non-keyword (ie click here) anchor text.

I assumed that the nofollowed page would not show up, but I included the page as an extra control measurement.

The results

About 3 months after the test was set up and the rankings didn’t change in two months, it was time to end the test. A search for [KW1 KW2] made Google only list two URLs in the results. Only page 1 and page 3 showed up. After clicking on ‘repeat the search with the omitted results included’, three more pages showed up in an order that was slightly different from what I expected. The optimized title tag link (#2 in the original set-up) ranked above the partially optimized anchor text link (#5 in the original set-up). As expected, the page where the nofollowed link pointed to, didn’t show up.

  1. An image link with a [KW1 KW2] alt attribute, but without a title tag.
  2. An image link without an alt attribute, but with a [KW1 KW2] title tag.
  3. A text link with a partially optimized (more about [KW1] here) anchor text.
  4. An image link without an alt attribute or title tag.
  5. A text link with a non-keyword (ie click here) anchor text.

The control test did not show any different results.

The conclusion

The conclusion of this particular test, is that an image link with an optimized anchor text is more valuable than a text link with an anchor text that is not optimized very well, in terms of ability to pass relevance.

An image link with an optimized anchor text might be more valuable than a text link with an anchor text that is not optimized very well. Because of the small difference in rankings and because of the order the links were placed in, a new test should be set up in order to

While lots of folks try to avoid image links (image based navigation included), this is absolutely not necessary. Depending on the anchor text possibilities, I would even prefer image links over text links in some cases. When you have an imaged based navigation with links to ‘Home’, ‘Services’, ‘Products’ and ‘About’, for example, you can add extra value by making the alt attributes more descriptive. This test shows that an optimized alt attribute (‘SEO Services’) might be better than a not so optimized anchor text (‘Services’).

The side notes

Setting up an SEO test that is 100% valid remains extremely difficult for several reasons. I probably made a small mistake somewhere while setting up this test, but I do think the results are pretty valid.

Please note that this test only included links to internal pages. It might be that image links that point to external pages are being treated differently by the major search engines. Also, the image format that I used was not a standard format. Google might value a link from an image of 70×15 px different than a link from an image of 468×60 or 250×250 px. This however, can (and will) be tested in a different set-up.

11 Responses to “A Good Alt is Better Than a Bad Anchor”

  1. Jeremy Rivera March 11, 2009 at 9:29 pm #

    Was this website listed in webmaster tools?
    Were the recognized inbound links recognized in webmaster tools affected/counted?

  2. Whitespark March 12, 2009 at 3:58 am #

    Great test! I have often chosen to use an image for links specifically so that I could utilize better alt text, and I have always wondered if I would be better off with a non-optimized text link. Now I know!

  3. Luc March 12, 2009 at 10:53 am #

    Hi Wiep. I think there is no difference for internal or external links.

    I have a website that ranks on the first page in google with a very competitive keyword (around 10 million results). There are nearly none textlinks with just this keyword. But there are hundreds of image links where the alt tag has this keyword. The ranking for this keyword is stable for about a year now.

    I created I widget (an image) that visitors can use on their own website. Users can copy the optimized code from my website.

    Now I try to rank well on anothor very very competitive keyword (67 million results) with the same technique. Since a few months, the site ranks on the second page in google.

    The website is 1,5 year old and has a pagerank of 5. I have to conclude that optimized image links are well appreciated by Google ;)

    If you need more info, please send me an email. I don’t want to disclose the website….

  4. Wiep March 12, 2009 at 11:35 am #

    @Jimmy: No, the website was not listed in G’s Webmaster Tools.

    @Whitespark: Thanks :)

    @Luc: Although image links that point to external pages surely do pass value, I do expect Google to treat these links differently, just like they do with “regular” internal vs. external links. It is quite common to use image links (ie navigation) to other pages on your own website, but -besides banner ads- it’s less common to do this to external pages.

  5. Erdal Gul March 12, 2009 at 3:13 pm #

    Hi Wiep,

    Its a nice testa altough its not a confirmation. Maybe this should be tested with several domains(existing and with new ones) with external links, other image formats. I will put this in my to do list. This is an interesting matter so let’s test it :-)

    Erdal

  6. Keonda March 14, 2009 at 5:07 pm #

    Hi Wiep,
    Very instructive test!
    Any clue on the ranking difference between an image link with an optimized anchor text and an image link using CSS image replacement?

  7. Michael Martinez March 20, 2009 at 6:06 pm #

    Good post. Good idea for a test.

    No test, no matter how valid its construction, provides a usable result if the test cannot be independently repeated and verified.

    Because it’s so difficult to describe these tests without linking to the test sites (and believe me when I say I understand why that is necessary on so many levels), we have to be careful not to read too much into anything.

    The greatest value of your test is in your sharing so that other people can attempt to construct similar tests. The more of these types of tests people create and share, the more likely that a consensus will form around the actual behaviors of the search engines rather than people’s individual experiences.

    A bad thunderstorm can literally alter the effects of any SEO test.

    1 test is just a start of a long process of study and sharing that, hopefully, leads to better understanding for all of us.

  8. Wiep March 23, 2009 at 3:56 pm #

    @Erdal: Agreed, further testing is always a wise thing to do.

    @Keonda: No, so that’s one of the factors that can be clarified by a new test.

    @Michael: I love to perform little tests like this and also like to read about them on other blogs. Let’s hope that it’s an inspiration for others to set up tests as well.

  9. Awanish May 11, 2009 at 8:36 am #

    Is splitting an Image is good practice for placing more keywords in alt tags.

  10. Wiep May 11, 2009 at 8:53 am #

    @Awanish I wouldn’t recommend doing that.

Trackbacks and Pingbacks

  1. en image alt tekst, wij gingen het voor u testen ! | Etail - April 21, 2009

    [...] Wiep deed een tijdje geleden ook een dergelijke test en schreef er zijn resultaten over in A good alt is better than a bad anchor [...]

Liked this post? Share it!

del.icio.usFacebookStumbleUponTwitter