NYTimes uses AI and statistics to study changes in Trump speeches

In short: Two New York Times journalists used AI and statistical checks to review about a decade of Donald Trump’s speeches and reported shifts in his speaking style.

What's going on

Peter Baker and Dylan Freedman at The New York Times analyzed public speeches by Donald Trump from roughly the past 10 years. They combined AI-assisted review with traditional reporting and basic statistics.

Their analysis found several changes over time. The speeches became longer and used more profanity. They also used more absolute wording, meaning more statements that sound like total certainty, such as framing things as always or never. The journalists also described the later speeches as less coherent, meaning the ideas were less clearly connected.

Freedman used code and AI prompts to surface examples of different patterns, like negative language and unusually long statements. You can think of this like using a smart highlighter to quickly point to possible examples. Baker and Freedman then checked those excerpts in full context and decided what to include in the story. They also used word counts and vocabulary measures, which are simple ways to track how often words appear and how varied the language is.

The Times stressed that this is an analysis of rhetoric, meaning speaking style and word choice. It is not a medical diagnosis and it should not be read as proof of cognitive decline.

What to watch

This kind of work shows how newsrooms may use AI to find leads faster, while still relying on people to verify meaning and provide context. Readers should watch for clear explanations of methods and limits, especially when AI is used to analyze public figures.

Source: NYTimes