Video summary

The AI Writing Epidemic

Main summary

Key takeaways

News and Commentary

Overview

The video argues that AI-generated writing is spreading online and making content feel “sterile,” homogeneous, and less creative—while also becoming increasingly detectable once you learn its patterns.

Key Claims

AI is reshaping writing faster than other industries

The creator cites research and surveys suggesting generative AI use is concentrated in content creation and editing, including:

  • A reported 48% of writers using AI (BookBub survey, 2025)
  • Findings attributed to Harvard Business Review (2024)

Some viral posts appear AI-written

The creator claims a viral Substack post about Anthony Bourdain may have been written by AI. The claim isn’t that it’s impossible for a human to write similarly, but that many people don’t recognize AI’s linguistic fingerprints—and that the post’s engagement metrics and style are consistent with AI output.

“Dead Giveaways” for AI Writing

The creator points to recurring features often associated with AI-generated text:

  • Negation flips and contrasts Example pattern: “it’s not X, it’s Y.” The creator argues LLMs frequently produce this kind of structure.

  • Frequent or awkward em dashes (—) The creator says humans sometimes use em dashes, but AI often uses them unnecessarily or mechanically for emphasis and structure.

  • “Rule of threes” phrasing Example phrases: “too raw, too weird, too true,” or other three-part lists. The creator argues humans vary these more naturally, while AI tends to produce neatly rhythmic, robotic triplets.

  • Specific punctuation inconsistencies In the Bourdain-related post, the creator claims em dashes were replaced with hyphens (possibly to disguise AI usage), and then notes similar dash-and-triplet patterns appearing again.

Why AI Writing Can Sound the Same (“Average and Safe”)

The video explains AI writing as a system that:

  • Predicts the next likely text token based on training data.
  • Is trained on large volumes of existing internet and book text, which encourages it to reproduce common statistical patterns rather than developing a genuine “internal world.”

Negative feedback loop

The creator warns of a cycle where:

  1. More AI output appears online.
  2. Future models train on AI-produced text.
  3. This can increase bias and further reduce creativity.

“Model collapse”

The video references “model collapse” (citing a 2023 Nature paper), describing degradation—like losing nuance or accuracy—when models are trained more heavily on AI-generated content.

Nuance: Limited Use vs. Outsourcing Voice

The creator distinguishes between helpful editing and outsourcing authorship:

  • Acceptable use: AI as an editing assistant—suggesting phrasing while keeping the author’s voice.
  • Warning: Don’t let AI “write the heavy lifting.”

Personal experience is included: the creator previously used AI to draft script parts due to workload/algorithm pressures, but now refuses after noticing it affected their voice.

Call to Action / Stance

The creator urges viewers to:

  • Spot and call out AI-written content
  • Identify patterns and filler language
  • “Distance” from AI in creative work

They emphasize that human strengths—especially creativity and improvisation—are at risk of being undermined when AI starts driving the development of a unique writing voice.

Closing Promotion

The creator promotes their writing project:

  • Novel: “Fractured Worlds” A story about a 21-year-old who can shift into parallel reality in a post-pandemic world.

  • Updates: via a Substack newsletter

Sources Referenced

  • Harvard Business Review (2024)
  • BookBub survey (2025)
  • Nature (2023) paper on model collapse

Original video