AI Detective: How I Spotted Clues of AI Writing in a Comedy Article That’s Killing It With Readers

🌈 Abstract

The article discusses the author's experience in identifying AI-generated content, particularly on the online publishing platform Medium. It explores the author's background in critical reading and textual forensic analysis, which has helped them develop a keen sense for spotting AI-generated text. The article also touches on the history of AI in copywriting and the influence of early pioneers like Amanda Weston and Sean Volsler.

🙋 Q&A

[01] Identifying AI-Generated Content

1. What are some of the key skills the author has developed in identifying AI-generated content?

• The author has a background in teaching critical reading at college and studying textual forensic analysis, which has helped them develop a sense for spotting AI-generated content.
• The author has also gained extensive exposure to vast amounts of AI-generated text, which has contributed to their ability to recognize it.

2. How does the author's experience compare to AI detectors in identifying AI-generated content?

• Research studies have shown that human experts sometimes outperform AI detectors in identifying AI-generated content.
• The author considers themselves an "old sea-dog" who can spot AI-generated text from leagues away, even before the widespread popularity of ChatGPT.

3. What are some examples of readers being unaware that the content they encountered was AI-generated?

• The author knows of readers who have gotten tattoos of phrases they saw online, without realizing the content was AI-generated.
• The author has also observed instances where readers have gullibly swallowed and acclaimed AI-generated content as the work of human authors.

[02] Analyzing a Suspected AI-Generated Article

1. What are some of the key characteristics the author identifies in the suspected AI-generated article?

• The article contains various "AI clichés" and rhetorical devices that the author has previously written about as being common in AI-generated text.
• The article follows a predictable approach with clear section headers and a linear progression, which the author attributes to the way language models function.
• The author notes the article has significantly outperformed the author's other genuine stories, which the author sees as a red flag.

2. How does the author demonstrate the similarities between the suspected AI-generated article and their own AI-generated content?

• The author provides an example of a custom-generated "Instant Satire" piece, which they claim has a similar style and quality to the suspected AI-generated article.
• The author suggests that readers can try to identify the original article themselves, as the similarities between the two pieces should be apparent.

3. What does the author's experiment with AI-generated content demonstrate about reader perceptions?

• The author suggests that when readers are unaware of the content's artificial origins, they may actually enjoy and prefer the AI-generated content over "organic" human-written content.
• The author argues that this is because AI-generated content is engineered to be palatable and popular, playing it safe to appeal to the crowd.