This article focuses on online assessments of a controversial British television advertisement. Across blogs, websites and forums, a range of stances emerged in debates about its possible 'sexism ' , 'humour' , 'misogyny' , or 'realism' . The analytic interest here is in the ways that assessors invoked 'f-' categories (e.g. 'feminism' , 'feminist') as part of their assessments: across the data corpus, people would locate themselves or others in relation to 'feminism' , where 'feminism' was variously old-fashioned, modern, prejudiced, vital, dogmatic, complex and/or many other things besides. To account for this variability, the article pursues an ethnomethodologically oriented policy of treating categories not as vectors for in-the-head social attitudes, but as resources for on-the-screen social actions. Categories thus became analysable not for what they revealed about their authors' real thoughts vis-à-vis feminism, but for how they functioned as crucial components of recipient-designed online assessments. Studying examples of positive and negative assessments, the paper subsequently shows that and how users claimed or denied their own (or some others') allegiance to 'f-' categories as a method for strengthening their own (or undermining others') assessments. A concluding discussion considers the wider applications of a categorial approach to feminism in a world of increasingly mediated interaction.