The Subcorpus Campanadas Gold Standard is a curated and statistically balanced dataset of Spanish-language social media posts published on X (formerly Twitter) in reaction to the annual Campanadas New Year’s Eve television broadcasts in Spain. The corpus is designed to support research in digital discourse analysis, sociolinguistics, and Critical Discourse Studies, with a particular focus on public evaluation, gendered judgment, and delegitimation processes in high-visibility media events. The dataset has been constructed using a stratified sampling methodology to counteract algorithmic bias toward viral content. Posts are grouped by target public figure and sampled under a strict equality strategy, ensuring comparable representation across highly visible and less prominent figures. Within each target group, data are further stratified by engagement level, distinguishing between high-engagement (viral) and low-engagement (organic, long-tail) discourse. This design allows researchers to analyze both dominant narratives and less visible evaluative patterns. All entries are manually annotated following the Judgment subsystem of Appraisal Theory (Martin & White), with labels covering evaluative polarity (legitimation vs. delegitimation), judgment category (Propriety, Capacity, Normality, Veracity, Tenacity), and salience strategies, including femininity-invoking versus norm-neutral evaluations. Annotations were produced as part of a gold-standard workflow intended for qualitative analysis, corpus-assisted discourse studies, and evaluation or training of computational models. The corpus is distributed in structured JSON format, preserving original post text, engagement metadata, target attribution, and expert annotation layers. The dataset is intended for reuse in academic research, methodological replication, and interdisciplinary studies of media rituals, online evaluation, and gendered public discourse in Spanish-language contexts.