Google Adult Images

How does Google decide what is suitable for general audience viewing. The opposite problem is also encountered, how do images that are not linked with adult subjects make it into the Google images only when safe-search is turned off?

Matt Cutts worked on Safesearch. It is an algorithmic process. From the Google blog:

Google’s filter uses advanced proprietary technology that checks keywords and phrases, URLs and Open Directory categories.

It is likely that factors used in SafeSearch would also be useful in determining what is defined to be an explicit image?

Despite the advances in computer algorithms to be able to recognise faces in images, there is still no real way for a computer to tell what is in an image, let alone what comprises an explicit image. What’s it going to do? Look at the amount of pink in the image?

In this video, Matt Cutts talks about images and adding alt information to describe the image. It is logical then to assume that this is this one of the key factors in determining which images get removed by safe search.Any naughty alt information is likely to send that image to the safe search off category.

Even without this information, other possible ways to tell whether images contain explicit content are the links to the images. If they come from sites which have adult content then this will also hint at adult related content.

The file name is likely to play some role. Some webmasters have suggested that the directory may even have an effect. If you have a mixture of safe and adult content then you might consider having two image directories. One for the safe images, website.com/images/ and one for adult images, website.com/adult-images/ just to make it clear.

It may be that text associated with the image is also a factor. What ever the exact method it is clear that it would err on the side of caution. It would be better for the site to have a non-adult image in the adult index rather than an adult image in the non-adult image.

Images are spidered by a separate robot, Google Imagebot.  It would seem,  on the face of it,  an easy experiment to test these factors, but the frequency at which the Google Imagebot comes around (from 3 months to a year) means the results would be a long time coming. Looking at the image results, Seothegame only has a few images spidered by Google images.Perhaps on a more popular website the data might come in faster.

There is also a small reliance on human intervention with the Google Image Labelling game, in which you and your unknown online partner spar to name images. The ones that you both choose are taken to be indicators of the content of an image. This is a good idea but it can only hope to label a tiny fraction of  the ambiguous material on the web.

If you are experiencing any problems with Google safe search, for example your non-adult images on your website have been included in the adult index than you can tell Google using this form.