At the time of writing image search is still in its infancy. The problem is that an image file is very different from a text document from the perspective of a computer scientist. The technology problems of searching for relevant text in a corpus of all text has largely already been solved.

Images are a completely different story for two primary reasons. First, image files are on average much larger than text files. Whereas a text document may be made up of thousands of 0s and 1s on the binary level, an image with the same dimensions may be made up of hundreds of millions of binary characters. This means that the computing power alone can become a bottleneck to solving the image search problem.

Secondly, algorithmically determining the meaning and contents of an image file is much more difficult than doing the same for a text file.

Whereas a word may have only a few meanings and contain very little, an image can have millions of meanings and contain many objects. Imagine the difference between the word “crowd” and an image of a crowd.

Although the word maps to the idea of a large group of people, the details of those people are not represented. This is not the same for an image of a crowd where the details of each person in the crowd are visible.

This of course is an oversimplification of the image search problem but it helps to illustrate its complexity. As a result of the complexity of these two factors (size and meaning/contents), search engines have relied on textbased metrics to rank images.

To make matters worse, these technical issues fail to address the social implications of image search. Whereas most search queries are done with the intention of getting information, the queries where images are the best result tend to aid people in stealing images. From an SEO perspective, this leads to the question of whether or not image search–referred traffic is even useful

Is Image Search Traffic Useful?

This is the most common question I get when talking to clients about image search. They are right in that many image searches result in the user simply stealing the image and not giving credit to the owner or the site where it came from.

My initial answer to this question is two-pronged:
1. I have never seen an example where image search referrals provide enough value that it warrants making image optimization worthwhile.

2 . At the same time, I think that everyone, regardless of limitation or disability, should be able to use the Internet. This includes humans and computers alike.

In order for this to happen, there needs to be a way to represent images for those who are visually impaired and those systems that can’t understand them (which includes search engines). The best way to do that today is to optimize images just like SEOs do for a website as a whole.

Alt Attribute (Alt Text)

The code to include an image file in HTML takes the following format:

<img src=”http://www.example.com/image-filename.png” alt=”Alt Text” />

The parameter in the alt attribute (in this example, Alt Text) is displayed in the event that the image cannot be represented. This happens when the technology accessing the tag cannot display the image. This is most common when the image is broken or when it is accessed by software like a screen reader.

This text typically describes the image and is a very good ranking factor.

Surrounding Text

Many times the text that surrounds an image in a web document describes the image. This can take the shape of a caption or introduction to the image.

This information is very useful for image search engineers. It is relatively common for an image to rank for a query that matches the text around an image.

Filename

In theory a filename should always describe its content. Unfortunately, this is not the normal case for image files. Many times they will be named by a computer (like a digital camera) rather than a human. This makes this a relatively poor metric for rankings.

That said, it does help to include descriptive words in image filenames because it is still used as a minor factor in image search rankings.

Traditional On-Page SEO Factors for Images

It is important to note that the search engines still use normal ranking metrics for images (relevancy and popularity). This is usually done on the page where the image is located. (Note this is different than the image URL.)

This means that by optimizing a page in the normal ways, you are also optimizing its images. It is a win-win situation.