I created TrumpLexicon when I wanted to know if Trump’s language was changing over time, like a dementia patient’s language might. And was all this talk about men’s asses related to any other trend. (If you’re reading this after 2026 that might sound odd, but there was a LOT of talk of how attractive Trump found other men’s butts in 2026-10.)
Nothing Trump posts to his Truth Social blog is included. Nothing Trump writes is included. No scripted remarks are included. Just what Trump says is being indexed. (The White House stopped posting written transcripts in 2025.)
Most transcripts come from Roll Call’s Factba.se, which publishes nearly everything he says on camera. Two official sources fill in around it: the government’s Daily Compilation of Presidential Documents (via GovInfo), and the American Presidency Project at UC Santa Barbara. Some sources respond poorly to requests from cloud servers. So there’s an unlisted page that lets my own computer fetch pages through a small Chrome extension and hand them to the server.
A Cloudflare Worker scrapes and serves. D1 database holds counts. R2 stores raw pages. Every quote on the site links back to its transcript and, where there’s video, to the moment in it. The same speech often appears in multiple sources. Each appearance is counted once, from the source that best captures his words, and the others are kept and marked as duplicates. Everything on TrumpLexicon ends yesterday as any partial day looks like a downward trend.
Adjustable date ranges. Heat maps. New words. Nicknames. Insults. Topic trends. Anything can be hovered over for more insight.
Analysis includes readability (Flesch-Kincaid grade level, Gunning fog, SMOG). Vocabulary variety, repetition, superlatives. Sentiment measured using VADER. Comparison with other presidents back to Carter.
Given the 2026-10 talk about men’s behinds, I also included context trends. This is because ass-talk used to be about firing, kicking, saving or suing before 2026-10’s admiring.
A related tool counts how often two words share a sentence. On a year and a half of data, “beautiful” shared a sentence with “Melania” about 10 times out of ~4,000 uses, far less often than with “bill”, “coal” or “ballroom”.
Let me know if you can extract any meaningful insights. My own learning was that many of my Trump-is-going-senile assumptions couldn’t be backed up by data. He has some downward dementia seeming trends but they’ve always bounced back. At least that’s what it looks like as of today.
