Hello, I try to answer your question and my answer is open to any comments and critiques. Please chek them, if there is any to ensure that my answer would not be misleading.

Content analysis is a very general term which refers to various methods and theories methods are based on. It can even refers to any analysis of content, sometimes regardless of its from (written or visual), that aims to find out characteriestics of content. It can basically classified into quantitative and qualitative ones.

In your case, you mentioned two terms, discourse and codebook. As far as I know, discourse analysis seems to be a qualitative one. Therefore, your mention of codebook is a little bit confusing. Since I am not sure about which kind of method (quantitative/qualitative) you plan to use for the content analysis, I can only provide a solution based on my experience, which is only quantitative one.

The nature or key logic of quantitative content analysis is to deductively identify features of content (which time are superficial most of the time, such as frequency of word, length of text, source, etc.) that are a priori in existing literature. This is why you need a codebook for your analysis.

Therefore, I would guess that, if you tend to conduct a quantitative content analysis, which is quite similar with framing analysis that I did before, you might follow the steps as follows:

1. Sampling and decide the unit of analysis. Check some textbooks of social science research method, or those about communication studies (in which I guess you major, thus I will recommend at the end) and decide which strategy of sampling that suits your research objectives and that is within the scope of your condition. Also, decide the unit of analysis, it can be a news item or a whole web page, which also depends on your objectives. The decision of unit of analysis is relevant to your question about webpage, make sure that you find a suitable one. For example, since there are too much information in which there are much that you do not need. You can make each single news item or one twitt, if you analyze social media. Also, I think the form of news item (newspaper, broadcasting, TV, social media, etc.) has nothing to do with your analysis because all data that you need is written instead of visual or audio.

2. Identify categories, which refers to features of content and in your case, from existing literature or propose some categories yourself based on the requirement of your research questions and identify definition of each category, which is the basis of your codebook.

3. Make a codebook with the definitions that you settle beforehand. The function of codebook is to help you find categories (it can be topics, or frame in news coverage).

4. Conduct conduct inter-coder reliability tests until the result is applicable. This step is to make sure the objectivity and replicability of your study, in other words, anyone would get the same or very similar results if they used your codebook.

5. Start coding, namely finding the categories based on the definitions in the codebook and if the definition is not clear or there is new categories, revise the codebook, conduct inter-coder reliability tests until there's acceptable result and start the coding again with the revised codebook.

6. Now you have your data, namely the number of each category in each unit of analysis. As you said your objective is to find the evident ones, I would guess that you need to calculate the frequencies of the categories in all your sample and figure out the ones with the frequencies you need.

7. You mentioned that you want to find 'moral' discourse, you might need to identify some value judgements from the content. However, it is very hard to do so through a quantitative content analysis because you can never know how the value judgements are expressed before coding. In this case, you cannot pose a definition of such value-laiden categories or your definition would be vague and thus hard for practice. Please do pay attention to this issue and add some qualitative analysis to find the deeper meanings with value in the content where you identified categories.