Or so Google Imagebot might have us believe. As a fellow contributor, participant, follower, and fan, I’m going to take Imagebot’s opinion with a large grain of salt — or a block of salt.
Let me explain. Recently I was doing a little image research and came across some images that grabbed my interest, especially given the growth in blended search, although image infiltration with blended results still seems rather minor. So what follows is a bit of that exploration — which is to say that it is a very small sampling, anything but exhaustive or even remotely scientific, and aims to drive discussion and even more exploration than anything else.
I came across this image doing image searches within Google:
What struck me as being odd was that it led me to SEOmoz, specifically to this post. This itself seemed a bit odd, but even more so when the source of the image was listed in the image results as “www.tobacofreeutah.org.”
Let’s dig deeper. Upon examining that post, we do not see the image anywhere. Looking closer, we find that it is linked to from the comments by member “carfeu” about 4/5 of the way down the page with the rather un-descriptive anchor text, [this]. Clearly not a lot of signaling going on here.
Playing around with different image searches in Google, mixing [seomoz] with various other bits of text from that blog post and comments, we continue to see this “clueless” image make an appearance, without any regard to proximity of the text to the image or related text. Although, if we search specifically for [seomoz clueless], then it shows up front and center. Also note that the word [clueless] does not appear anywhere on the page.
What’s particularly interesting though is that this image doesn’t actually appear on SEOmoz, but has only been linked to with a single link from a blog post comment. Furthermore, it has been linked to with a single stop-word anchor text, [this]. Actually, doing an image search for [seomoz this] doesn’t reveal this image until the 12th page of results in Google image search.
But searching on [seomoz] combined with other sampled text from the page, things like [seomoz roundup], [seomoz jane], [seomoz rebecca], [seomoz identity], [seomoz sean], etc., brings up the image, generally, within the first 6 pages of results.
What is fascinating here is Google’s rather broad association when it comes to images. In fact, possibly too broad. Now, if you search within the regular web search in Google, even on [seomoz clueless], no images even come up. But if you search for [seomoz clueless image], then we get our image and a couple of blog posts — though neither of them are the one from which the image is referenced. And to make matters even more interesting, change the order to [seomoz image clueless], and once again we get no images.
And how does Yahoo and MSN fare on these searches? Well, apparently, neither of them consider SEOmoz to be clueless at all and don’t return any images for [seomoz clueless] at Yahoo or [seomoz clueless] at MSN, nor did they return this same image for other words on the page like Google did.
Depending on your perspective, you might consider Google to be much better at handling image searches, or much worse and a bit of a catch-all. Google seemed to connect the images, even those only linked to, with words on the page, regardless of whether those words had any correlation to the image, such as file name, hosted domain, etc.
Yahoo, on the other hand, seemed to be almost exclusively returning those images that had been tagged in Flickr with the text, or had the text in the image name — I had to really dig deep into the results to find exceptions, and those seemed to at least require a relation of proximity or prominence.
MSN actually seemed to carry things a step beyond the file naming and made additional connections to these, such as returning images of Brittany Murphy (though fairly deep in the results) for the image search on [clueless]. The Brittany Murphy result was interesting because there was no reference to the word [clueless] on the page where the image was located, but she did appear in the movie Clueless.
So what’s it all mean? Well, as stated in the beginning, this was rather surface level, and after awhile, quite easy to get turned around in circles, but an interesting exploration in the area of image searches. It’s by no means definitive, but some takeaways and possibly a confirmation of some concepts are:
- keyword-rich file naming for images is especially important for Yahoo and MSN,
- though not necessarily as important or even required in Google.
- Google, however, is extremely generous in associating images with keywords, possibly even without regard to prominence or proximity to the image itself.
- Not only that, Google makes an association to images simply through links to the image.
For even more interesting fun, try searching without [seomoz] but with other words from that blog post paired up with [clueless] and you’ll also come across this image referenced by the post on SEOmoz. With this rather broad association, it seems like Google image-bombing wouldn’t be too difficult — not that I’m trying to encourage that.
What I am trying to emphasize is that some careful or creative image optimization may go a long way. And while not even touched on, perhaps the other interesting aspect here to our clueless image is that the link to it is also attributed with a “nofollow” — I’ll leave that for someone else to explore.
And no worry Mozzers — I don’t care what Google Imagebot may think, SEOmoz is anything but clueless and has the membership that will state otherwise.