Zemanta is like having a friend looking over your shoulder while you write, making suggestions. It analysis your text as you write. Then it delves into a database finding images tagged with the words you use.
At the same time, it searches web sites looking for similar material related to the words. Finally, it suggests Technorati tags.
This sounds great. And in theory it could be. The problem is, if Zemanta was your friend, he or she would quickly wear out their welcome and quickly become an irritant. That’s because there’s something wrong with the suggestions. While the problem is across the board, it is worse with images.
Zemanta suggests dozens of images. Few, if any, are usable. Few are vaguely relevant. Many are surreal.
This is my 51st blogpost. In my earlier 50, Zemanta managed six plausible images choices. Two were logos I could have found myself. One was a generic shot that I used out of desperation. So, in round numbers Zemanta finds an acceptable image for one story in ten.
I suspect Zemanta only has access to a limited database because the same pictures kept appearing in the suggestions — although to be fair, I cover a narrow subject that doesn’t easily lend itself to illustration.
These problems are not as bad for links.
Zemanta suggests in-story links for certain words and after-story links to related items. The in-story links are mainly to Wikipedia or to company sites.
For example, mention IBM and Zemanta suggests both the Wikipedia entry for IBM and the company’s home page. This isn’t useful unless there’s a benefit in providing lots of outgoing links that I’m not aware of. The news links aren’t too bad, but many are out of date.
Zemanta claims its software “makes blogging fun again”. Well maybe it does people, but in my experience it makes blogging more frustrating than it needs to be. Maybe it will improve in a future incarnation. I hope so, because the idea behind Zemanta is good, sadly the current implementation and the data it has access to lets things down from my point of view.