Rankmaniac 2012

Saturday, February 4, 2012

The Pitfalls of Writing a Clickbot

One way to up your image's pagerank (and win Rankmaniac 2012) is to get your link clicked on a lot. Therefore, it would make sense to write a clickbot that automatically searches for your link and clicks on it. I had several ideas for writing a click bot.

First, I wanted to write a python script that fetches images.google.com, looks for the input field, performs an image search and then clicks on our image. This is the simplest way to write a click bot and win Rankmaniac 2012. There are some downsides. First, Google randomizes the id, name, etc... of their input field and of their other elements on the image search page. However, I could probably just search the page for the input field with the correct attributes and perform the search with that element. Another downside is that google does a pretty good job about detecting click fraud. Here's a study that Google did about it (http://www.usenix.org/event/hotbots07/tech/full_papers/daswani/daswani.pdf).

My new idea is to use some sort of Custom Search API to write a bunch of search queries in javascript, put the pages on line and run them through proxies to simulate a lot of diverse traffic. However, Google doesn't really have a good search API and this would be difficult. Also, they'd still probably be able to detect the click fraud. My search continues...

Friday, February 3, 2012

How Google Ranks Images

We're a few days into the Rankmaniac 2012 competition now and what we've learned is that we pretty much have no idea what we are doing. We attempted to harness the power of social networks and hilarious memes, but the only images of ours that show up on Google image search are the ones that have been on the web for a while and were popular before we posted them - the Super Bowl image, the Google and Bing logos, and my Google+ profile picture (this one still baffles me).

Why are these images doing better than our own original content? The first is that we linked directly to the url of this image (our blog doesn't host these). Since these images already appear all over the web, Google has deemed them the most important pictures on our blog (again, my profile picture is still an conundrum). So, even though none of these images are the ones we are trying to rank, they do pretty well in the search.

So, to actually learn something about search engine optimization (SEO), we need to figure out how to rank our own images on image search. Bryan and Judy have been posting some pretty adorable and kind of funny images but none of them have made it to the search. By googling "how google ranks images", we learned that there are a few image factors that affect the image search results:

image filename: All of the pictures we have posted (except those we linked to) have in their filenames "rankmaniac2012" or some variation of the query keyword.
text around the image: Descriptive captions are a plus! We have tried to include the keywords anywhere we can near the images and all over the blog.
alt and title attributes: The alternative text of the image (the words that show up instead of the picture if it doesn't load) and the title of the image (the words that show up when you scroll over the picture) give Google more information about what the picture is and, when related to the keywords, will give it higher rank for the query "rankmaniac 2012".
PageRank of the website: This is where we are lacking. We need more websites to link to us!
Focus of the webpage: Our website is about the rankmaniac 2012 competition, so it needs to look like a legitimate webpage with relevant content.

These are the things that we intend to try for the duration of the contest, but if anyone has other suggestions we would much appreciate it ;).