My last post, Google Image Traffic and Flickr, talked about my experience hosting images on Flickr and the total lack of traffic I received from Google Image search. I decided to take it a step further and see if I can get some more data on this.

My general hypothesis is that Google gives preference to images hosted on the same domain as the page it is linked from.  

The rationale for this from Google's standpoint makes sense. If they didn't do this, then people who link to images from other domains would possibly, and in many cases quite probably, get a higher SERP for someone else's content. It is like the problem with scraper sites and RSS, but in the case of images it is much easier to fix because the links are usually obvious and live. The problem on Google's end can be solved by just giving priority to images hosted on the same domain.

To test this I ran searches on several common keywords and looked at the images which were on the first page for each search. I didn't bother with images beyond the first page because I'm looking at if there is a preference for same domain hosting. If it is found lower in search results, then I don't think it gives weight to disprove the claim.

First, some very common keywords:

George Bush (14.6m results) - 20 of 20 images were hosted on the same domain. There was a mix of mainstream newspapers, blogs, government, and education sites. No Flickr images.

China (230m results) - 20 of 20 images were hosted on the same second level domain. Several news sites had different third or fourth level domains for images, but the same second level domain. No Flickr images.

Paris Hilton (6.6m results)  - 13 of 20 images were hosted on the same domain. There was a Photobucket image ranked #2, but the link went back to Photobucket, not another website with a link. The top image was linked from a wordpress.com blog, but the image was located on indiewire.com. A celebritycrunch.com image was linked from the ABC in Australia. There doesn't appear to be any common thread behind most of these. No Flickr images.

Clouds (18.5m results)- 19 of 20 on the same domain. No Flickr images.

Next, some lesser keywords:

60 inch plasma tv  (711 results) - 19 of 20 were hosted on the same domain. No Flickr images.

tiny tim ukulele dr demento (168 results) - 20 of 20 hosted on the same domain. No Flickr images.

death valley california rocks sliding mystery racetrack (13.3k results) - 17 of 20 self hosted. No Flickr images.

Next, I went the images which rank highest currently on Flickr on their interesting scale. The purpose of this was to give heavy weight to the Flickr images with the most "buzz." I used words from the title of the image to see how they ranked in Google.

true love lingers (71k results) - A Flickr image was ranked #1. It was not the image that came up on the interestingness list, however. :)

Rhododendron against a White Wall (40.6k results) - 19 of 20 images were from the same domain. The #1 and #3 images were from Flickr. The #3 image came from a non-Flickr link. The images on the Flicker interestingness page did not show up.

I then tested some images from popular photographers on Flickr. Rebekka Guðleifsdóttir was named the top photographer on Flickr. I search for some of her images. She has 1,242 contacts. She has 10 images with 1,000 or more Flickr faves (I only looked for titles in English).

Bad Face Day - This image ranked #10 for "bad face day".

and then the bridesmaid just took off (88.5k results) - Ranked #1, #2, and #3. The image appears in different places on Flickr.  The #3 image is scraped.

My conclusion 

Google, for good reason, will give more "image juice" to images which are found on the same domain they are linked from. A much more extensive test could be run with a good script, but everything I found here conforms with what I've seen before. 

It is possible to rank on the first page for a search when an image doesn't come from the same domain, but the odds are against you. Even on the Paris Hilton search, the majority were self hosted.

It is possible to rank high with Flickr images, but odds are the link will go directly to Flickr, not to the site which links the image.  You will probably also need a very specific title to get ranked.

If you are a photographer, you can get many views from within Flickr. I've had over 20,000 views of my images and I know people with several time more than that. However, it does little to drive traffic to your website. It is all contained in the Flickr universe.  

Flickr was never designed to be a social networking site, despite many of the similarities it has with other sites. If you are interested in SEO or driving traffic, Flickr is probably a poor solution to your image hosting needs.  It is great, however, if you just want to host images somewhere and not worry about traffic or search engine rankings.