Case Study: Google Image Search and Marissa Mayer

SEO Case Study: Google Image Search Optimization

Experiment: To try to determine what factors play a role in optimizing for Google Image Search

Hypothesis: Image file names and alt tags play a role in determining image relevancy for a particular search term

Background: I was reading the Happy Birthday Gurtie post at threadwatch which led me over to Gray Hat News. Once on Gray Hat News I wound up in the “she’s a geek but her clothes match” thread about Google’s PR representative Marissa Mayer. I remembered reading the original business week article and finding that quote particularly funny. In the GHN post they have a satire about Marissa’s fashion sensibilities complete with pictures. So I said to myself “self if you were looking for pictures of Google’s Marissa Mayer where would you get them?” and I came up with the answer of Google Image Search. So I searched for [Marissa Mayer] and found some of the pictures, but not all of them. However the one picture that caught my eye was a group of cheerleaders. Why, because I remembered reading Marissa Mayer was a cheerleader from the business week article. So I wondered if this was a long lost picture of Marissa from her High School days. Upon examining the website I noticed this was not the case at all. However underneath the picture there were a series of names two of which were Marissa Carroll” and Kellie Mayer“. So clearly Google was using broad match text on the page to deliver images SERPS, not something I expected.

Case Study:
Looking at the image SERPS for [Marissa Mayer] the results are reasonably accurate. Some of the pictures (8) are indeed of Marissa Mayer (screenshot). Seven of the results have either [Marissa] or [Mayer] in the file name, six of which are of Marissa Mayer. The two additional pictures of Marissa Mayer actually have both [Marissa] and [Mayer] in close proximity to the image. Let’s look at the cheerleading page since that’s the one that piqued my curiosity. I’ve saved a copy of the HTML file lest it get pulled at some point.

When we strip out all of the formatting and HTML this is what we are left with

<img src=""
border="" alt="cheer002.jpg" vspace="0" hspace="0" width="300">
The Squad
Lauren Allen Claire Gittemeier
Marissa Carroll Katie Gorham
Kelsey Clark Kellie Mayer
Libby Coulter Meaghan O'Malley
Nikki Currie Brooklyn Presta
Kelsi Drummond Molly Smith

So taking a look at the page in a linear fashion we can see there is a reasonable proximity between the image the word [Marissa] and the word [Mayer]. In fact [Marissa] is within 42 characters of the end of the image tag and [Mayer] is within 91 characters of the image tag. So if its just proximity to the picture if we shift the names around and get the same picture in the image SERPS. [Lauren Gittemeier] works (screenshot), so does [Marissa Gorham] (screen shot) as well as [Kelsey Mayer]. Let’s see if there is a limit to how far apart the terms could be [Lauren Gorham] works , [Lauren Clark] however comes up blank after nearly 500 results. Now [Lauren] is 67 characters away from [clark] and we’ve established it can read farther into the linear text string for a match, so I suspect it’s the large number of other pictures that had better matching parameters. So stepping forward to the next name combination [Lauren Coulter] we again get a match, however [Lauren Meaghan] doesn’t give us the picture. The term [Meghan] is 95 characters from the end of [Lauren].

Now lets look at the 11 pictures that don’t have Marissa or Mayer in the filename and look at the proximity of both [Marissa] and [Mayer] to the image in question

IMG1: Marissa (1) Mayer(n/a)
IMG2: Marissa (n/a) Mayer(n/a)
IMG3: Marissa (n/a) Mayer(n/a)
IMG4: Marissa (1) Mayer(8)
IMG5: Marissa (-59) Mayer(-51)
IMG6: Marissa (1) Mayer(n/a)
IMG7: Marissa (8) Mayer(31)
IMG8: Marissa (42) Mayer(91)
IMG9: Marissa (-13) Mayer(n/a)
IMG10: source page unavailable
IMG11: Marissa (8) Mayer(16)

One page was unavailable and two pages had no mention of [Marissa] or [Mayer] anywhere on them. Looking at the content of those pages I’m willing to bet that at some point they did have the words [Marissa] and or [Mayer] on them, but I could be wrong. The overwhelming majority of the pages do have either [Marissa] or [Mayer] within 91 characters. It’s interesting to note that it can also read backwards from the image as well. I repeated the process on some other terms as well like [if you have small children] and [my sleeping dog] and had similar results, so I’ll spare you the disection.

If you are interested in optimizing for Google Image Search make sure you put the words in close proximity either before or after the image. Naming the files with the keywords seems to help, but is less of a factor than the surrounding text.

Special thanks go out to Google’s own Marissa Mayer for being a participant in the experiment, and to Gurtie of GHN Happy Birthday and thanks for sending me down this path. runs on the Genesis Framework

Genesis Framework

Genesis lets you to quickly and easily build amazing websites with WordPress. Whether you're a novice or advanced developer, Genesis provides the secure and search-engine-optimized foundation that takes WordPress to places you never thought it could go.
It's that simple - start using Genesis now!

Take advantage of the 6 default layout options, comprehensive SEO settings, rock-solid security, flexible theme options, cool custom widgets, custom design hooks, and a huge selection of child themes ("skins") that make your site look the way you want it to. With automatic theme updates and world-class support included, Genesis is the smart choice for your WordPress website or blog.

tla starter kit


  1. Text Link Ads - New customers can get $100 in free text links.
  2. - Get a premier listing in the internet's oldest directory.
  3. Need an SEO Audit for your website, look at my SEO Consulting Services
  4. TigerTech - Great Web Hosting service at a great price.
More in Case Study, Google, npb, npp, ntt, SEO
Constant Content Review

I first heard of through Aaron Wall's blog (selling content articles). I've linked to them once or twice but...

Blogging Tip : Permalink

If you write a blog or read them regularly chances are you come across the term permalink. A permalink is...

Blog News : Scrapes, Fakes, And Remakes

If you're a regular reader here you know I'm a big fan of integrating a trend or news blog into...