The concept of targeting social media users has received a great deal of attention following the Facebook and Cambridge Analytica scandal. Researchers have known for some time just how accurately this could be done using freely available data such as likes without even the additional data used by Cambridge Analytica.
If you read about the practice of user categorization, you might find reference to a 2013 study by Kosinski and colleagues. Some complain about public access to research data, but this study is freely available so you can take a look. Actually, research reports of this type are available to the public, but you would likely have to make the trip to the library on a nearby college campus.
Kosinski and colleagues establish the feasibility of categorization of Facebook users. They did this by asking for access to the the Facebook likes of 58,000 Facebook users and also additional information about these individuals (demographic data such as sex, marital status, political party affiliation) and information from personality inventories, etc. Once they have this data sets, the researchers are able to determine if the types of likes these individuals generate can be related to the other variables and how accurate one could predict these variables from the likes. This is a statistical process. The technique they built from the statistical data could differentiate homosexual and heterosexual men in 88% of cases, African Americans and Caucasian Americans in 95% of cases, and between Democrat and Republican in 85% of cases. So, if you wanted to send a specific message to Republicans or Democrats in the state of Iowa you could do so with a high degree of accuracy.