Delving into LLM-assisted writing in biomedical publications through excess vocabulary
Used for the PubMed abstract analysis and the rise of words such as "delves," "showcasing," and "underscores."
This page gathers the main outside references used for the article's claims, charts, and quizzes. The Wikipedia examples are named inside each expanded sign, while the broader studies and essay sources are collected here.
Research and data
Why Does ChatGPT "Delve" So Much?
Used for lexical overrepresentation in LLMs and the discussion of why words like "delve" become overused.
How much are LLMs changing the language of academic papers after ChatGPT?
Used for the multi-database tracking of LLM-associated terms across scholarly databases and PMC full text.
Comparing LLM-generated and human-authored news text using formal syntactic theory
Used for syntactic differences between human-authored New York Times-style writing and LLM-generated news text.
How Close is ChatGPT to Human Experts? Comparison Corpus, Evaluation, and Detection
Used for comparison between ChatGPT and human expert responses, plus detection work using the HC3 corpus.
The Uncanny Valley: An Empirical Study on Human Perceptions of AI-Generated Text and Images
Used for the uncanny-valley framing around human reactions to AI-generated text and images.
Towards Understanding Sycophancy in Language Models
Used for the explanation of sycophancy in AI assistants and how human preference feedback can reward agreement over truthfulness.
ELEPHANT: Measuring and understanding social sycophancy in LLMs
Used for the broader idea of social sycophancy, including LLMs preserving or affirming a user's self-image.
Chatbots Have Thoroughly Infiltrated Scientific Publishing
Used for the raw PubMed "delve" counts shown in the chart.
More Articles Are Now Created by AI Than Humans
Used for the estimate that AI-generated articles briefly surpassed human-written articles in Graphite's Common Crawl sample.
How Does AI-Generated Content Perform in Search and Answer Engines?
Used for the comparison between article volume, Google search results, and ChatGPT citations.
Rhetoric and examples
Why ChatGPT writes like that
Used for the article's discussion of antithesis, tricolon, parallelism, and taste in AI-like prose.
It's happening: People are starting to talk like ChatGPT
Used for the discussion of chatbot vocabulary bleeding into human speech and writing.
Jeremy Finds AI
Used for the extra section on AI-video narration, influencer speak, and default generated scripts.
New York Times - Signs of AI Writing
Used as the New York Times source note for the field guide.
The Simple Secret of Manhood Is Love
Used as a human-written quiz passage that still resembles common AI rhetorical patterns.
Why "Chrono Trigger" Holds Up in 2018: A Fresh Perspective
Used as a human-written quiz passage with AI-like journey and rule-of-three phrasing.
Atlas Obscura on Los Angeles food
Used as a human-written quiz passage with language that can feel AI-adjacent out of context.
Wikipedia: Signs of AI writing
Used as the core source for the Wikipedia-specific signs, examples, and field-guide framing.
See also: AI use
Conversations with AI used while developing and revising the field guide.