Each row is a word; each column is one of 95 Trump-account posts in date order, February 28 to April 18, 2026. A printed mark above a row’s line: that post uses the word. A cut below it: the share of that post’s own replies that use it (log scale, one scale throughout; a nick is under 1 in 1,000). Top eight: words his posts use far more than the replies. Middle eight: words the replies repeat under the posts that use them. Bottom eight: words the replies use and his posts rarely or never do. Words are chosen by a stated rule from the data, ignoring web addresses and @mentions; two profane words are masked. Presence only, not sentiment or stance.
Data: Muhammet Akkurt, Kaggle release 5, CC BY-NC 4.0; a curated 2026 sample. Repliers are not verified followers. Code: Luke Steuber, MIT.
Data