Alejandra Caraballo, writing for The Dissident:

Over the last few months, I utilized the Times’ API to pull the metadata for every single article relating to transgender issues since January 1st, 2014. This was done both through the Times’ own internal tag for transgender-related issues titled “transgender/transsexuals” as well as including a separate search through their articles API for a slew of trans-related terms that come up in coverage such as gender dysphoria, gender affirming care, and even hostile terms such as gender ideology. This resulted in a total of 3242 articles between January 2014 and early 2026. I subsequently utilized the metadata to be able to get the body text from archives in the Wayback Machine to be able to analyze the individual text of every article.

Once I had a corpus of all 3,242 articles written on transgender issues over a 12-year span, I was able to then run them through multiple LLM classifiers alongside a rubric to measure multiple dimensions of how each article framed its story and who it chose to center the story around. While many have reservations and reasonable concerns about the use of LLMs, language analysis is genuinely one of the useful use cases of this technology. Prior rules-based attempts such as natural language processing tools like VADER can’t capture the nuance within individual stories to measure and analyze thousands of stories and millions of words over time. Thus, I used three separate models, Claude Haiku 4.5, Gemini 2.5 Flash-Lite, and the open-source Qwen 2.5 32B to individually score each article on a series of measures. Each individual score may not completely capture the totality of an article or even all of the nuance. However, once the data is aggregated, clear trend lines emerge.

This is an incredible data visualization project, and a poignant reminder of just how far the New York Times as an institution has fallen. I canceled my subscription in 2019 after being fed up with their disgusting behavior, and refusal to acknowledge repeated factual, and morally reprehensible mistakes.