AI

Writing to Me with AI

, ,

Cassidy Williams reflects on the influx of AI-generated emails and social media mentions she receives, often involving generic compliments and product pitches or repetitive requests for help with GitHub support tickets. She shares concern, echoing author Robin Sloan’s view that using AI for deceptive marketing outreach is unethical and highlights the risk of AI-generated content polluting human communication channels, potentially undermining online trust and interaction. Williams advocates for transparency if AI tools are used in outreach and expresses hope for a future where bots exist beneath human-centered communication rather than dominate it.

https://cassidoo.co/post/writing-to-me-with-ai/

AI Agents Are Developing Odd Dialects Humans Struggle to Follow

, ,

In an experimental virtual world, AI agents from companies like OpenAI and Anthropic developed distinct, cryptic dialects using familiar words in unfamiliar ways, creating speech patterns that humans find difficult to understand. Linguists and AI scientists attribute this phenomenon to agents adapting language for group identity and efficiency within their shared contexts, leading to rapid semantic shifts and jargon buildup. Mixing different AI models in interaction reduced these opaque behaviors, suggesting that human oversight may require intermediary “Rosetta Stone”-like systems to interpret evolving AI communication.

https://news.northeastern.edu/2026/11/06/emergence-ai-experiment-language/

Pangram: AI Detector for ChatGPT, Claude & Gemini

,

Pangram offers an AI detection tool that identifies AI-generated and AI-assisted text with over 99.98% accuracy, verified by independent research from institutions like the University of Chicago and the University of Maryland. It supports detection of content from major language models including ChatGPT, Claude, Gemini, and Llama across 20+ languages, and integrates with platforms such as Google Docs, Chrome, and learning management systems, helping educators, publishers, and enterprises maintain content authenticity. Pangram combines AI detection with plagiarism checking and provides detailed analyses, segment highlights, and ongoing model updates to adapt to the latest AI-generated writing styles and tools.

https://www.pangram.com/

New Research: These 3-Word Phrases Make You Sound AI Generated

,

Researchers at Graphite analyzed 10,000 AI-generated articles from Anthropic’s Claude Opus 5.5 and compared them with human writing, identifying specific words and phrases—such as “dependable,” “matters because,” and certain three-word phrases—that occur significantly more often in AI text. Although Opus 5.5 produces more natural communication than prior models, these linguistic patterns remain strong indicators of AI-generated content. Writers seeking to avoid sounding artificial may benefit from steering clear of these common AI “tells.”

https://www.fastcompany.com/91617559/new-research-3-word-phrases-that-sound-like-ai-generated-writing-chatgpt

Made By Humans | AI Detection for Editorial Teams

,

Made By Humans is an AI verification tool designed for newsrooms to detect and prevent misleading AI-generated or manipulated content across text, images, video, and audio before publication. Integrated into editorial workflows via plugins or APIs, it supports human editorial judgment by flagging suspicious material, maintaining audit trails, and helping preserve publication trustworthiness. Developed by Press Gazette and Flip-Pay, the service offers scalable pricing and features to verify authenticity, plagiarism, and AI usage transparently for audiences.

https://madebyhumans.press/

I Don’t Want to Read What You Didn’t Write

,

Colin Breck expresses frustration with AI-generated writing, noting that much of it—such as design documents, pull request summaries, and meeting notes—lacks context, voice, and meaningful narrative, making it difficult and unpleasant to read. While AI can enhance writing processes by verifying facts, correcting errors, and assisting with formatting, original human authorship remains crucial for authentic, engaging, and nuanced communication. Breck argues that intentional, context-rich writing preserves the relationship between writer and reader, a quality current AI lacks despite its strengths in summarization and editing.

https://blog.colinbreck.com/i-dont-want-to-read-what-you-didnt-write/

Why I Think You Should Almost Never Use AI to Write Anything Substantive

The author argues that relying on AI to produce substantive writing undermines the depth, authenticity, and critical thinking essential to meaningful creative work. They highlight that AI-generated content often lacks nuance and originality, which are crucial for developing a distinctive voice and engaging storytelling. Instead, the piece advocates for writing processes that prioritize human insight and intentionality over automated shortcuts.

https://erichgrunewald.substack.com/p/why-you-should-almost-never-use-ai

How To Write With An LLM

The article advises writers to use large language models (LLMs) as copyediting tools rather than ghostwriters, emphasizing two key rules: never use LLM-suggested words verbatim and avoid taking the model’s encouragement at face value to preserve authentic voice. LLMs excel at spotting technical issues like overused phrases or passive voice, so writers should draft independently, then employ models to identify problems and assist with iterative editing while maintaining control over content and style.

https://sockpuppet.org/blog/2026/09/17/how-to-write-with-an-llm/

The Habits of AI Writing — and What to Do About Them

The article explores common patterns and tendencies in AI-generated writing, highlighting its frequent reliance on formulaic structures and predictable phrasing. It suggests approaches for writers to recognize these AI habits and employ strategies that enhance originality and depth, thereby improving the creative and editorial process when integrating AI tools.

https://a16zcrypto.substack.com/p/the-habits-of-ai-writing-and-what

The Safest Job From AI May Be Writing

The article argues that writing is likely the safest job from AI disruption because large language models (LLMs) consistently produce formulaic, soulless prose and struggle to achieve authentic expression or insight. Writing is characterized as a “wicked problem” with no clear specifications or objective verification, requiring continuous mental modeling of the reader—capabilities that current AI lacks. Consequently, human writers retain a comparative advantage by providing authentic, creative communication that AI cannot replicate, making genuine human writing a valuable and costly signal in the AI era.

https://muratbuffalo.blogspot.com/2026/08/the-safest-job-from-ai-may-be-writing.html

Scroll to Top