I am curious how comscore determines gender, and would love to see some official documentation or information regarding this if possible.
ComScore reports estimates of how much activity websites get, among other things, similar to Alexa and Quantcast. Their data comes from Internet users who presumably tell ComScore, among other things, what their gender is. It also sounds like the numbers may be adjusted to account for demographic estimates. Here are details from Wikipedia (click through for several documentation links):
comScore maintains a group of users who have monitoring software (with brands including PermissionResearch, OpinionSquare and VoiceFive Networks) installed on their computers. In exchange for joining the comScore research panels, users are presented with various benefits, including computer security software, Internet data storage, virus scanning and chances to win cash or prizes.
comScore is up-front about collecting user data and the software's ability to track all of a user's internet traffic, including normally secure (https://) connections used to communicate banking and other confidential information.
comScore estimates that two million users are part of the monitoring program. However, self-selected populations, no matter how large, may not be representative of the population as a whole. To obtain the most accurate data, comScore adjusts the statistics using weights to make sure that each population segment is adequately represented. To calculate these weights, comScore regularly recruits panelists using random digit dialing and other offline recruiting methods to accurately determine how many users are online, aggregated by geography, income and age. Correcting the comScore data requires having accurate demographics about the larger pool of users. However, some comScore users are recruited without being asked to give demographic information and, in other cases, users may not be truthful about their demographics. To ensure the accuracy of the data, comScore verifies its users' demographics during the course of measuring statistical data.
The corrected data is used to generate reports on topics ranging from web traffic to video streaming activity and consumer buying power.
I'm not familiar with comscore. But if you are asking how it identifies the gender writing style of a website, there are certain metrics that can be compared. For example, females tend to use pronouns more often and determiners less often than male writers. There are many other characteristics. Here's an article about gender writing styles.