This Is How to Tell if Writing Was Made by AI
Get the full intelligence
Search transcripts, export clips, track mentions, and explore all topics from “This Is How to Tell if Writing Was Made by AI” inside PodZeus.
In this episode of Odd Lots, hosts Jill Weisenthal and Tracy Alloway explore the growing challenge of identifying AI-generated text in an era where up to 40% of internet content may already be AI-written. The conversation centers on Pangram Labs, a company founded by Max Spiro that uses deep learning models to detect whether text was written by humans or AI. Spiro explains that while AI excels at grammar and clarity, it often fails in stylistic nuance and decision-making patterns, which Pangram’s model detects through millions of human-AI comparisons. The hosts express concern about the erosion of trust in written content—once signaled by proper punctuation and spelling, now easily faked by AI. They also discuss the ethical and societal implications, including reputational risks for writers, the rise of AI-driven content farms, and the need for new norms around transparency. The episode ends with a philosophical reflection on how the internet’s credibility is being undermined by an endless stream of 'cranks with good grammar,' forcing us to rethink how we evaluate truth and authenticity online.
Up to 40% of internet content is now AI-generated, especially in SEO-driven articles and social media posts.
AI detectors like Pangram Labs use deep learning to identify subtle decision patterns in writing, not just surface-level cues like punctuation.
False positive rates are low (1 in 10,000), but the risk of mislabeling human writing as AI remains a serious concern for creators.
The core problem isn't just AI writing—it's the severance of the old heuristic that good grammar = serious authorship.
New norms are needed: it's increasingly rude to share AI-generated responses without disclosure.
…and 2 more takeaways available in PodZeus
The Rise of AI Writing and the Trust Crisis
The hosts open with their personal unease about detecting AI-generated text, noting how even subtle stylistic cues can feel 'sickly' or 'overly earnest.' They question the reliability of traditional heuristics like grammar and spelling as indicators of intelligence or authenticity in writing.
Introducing Pangram Labs: The AI Detector
The hosts introduce Max Spiro, founder of Pangram Labs, a company that uses deep learning to detect AI-generated text. Spiro explains their methodology: training models on millions of human-AI text pairs to identify subtle decision patterns in phrasing and structure.
How AI Detection Works: Beyond Grammar
“Every sentence has dozens or hundreds of ways to phrase it. Over 50 or 100 words, you’re making thousands of decisions. If the vast majority line up with how an LLM writes, it’s vanishingly unlikely a human did it.”
The 40% Statistic and the AI Content Economy
“It's about 40% from an internet page perspective. About a year and a half ago, we looked at Medium and found that over 50% of newly written Medium articles were AI generated.”
The Ethics of Detection: False Positives and Intent
The hosts and Spiro discuss the dangers of false positives—when human writing is flagged as AI—and the ethical implications. They explore the difference between malicious AI spam and journalists using AI for efficiency, emphasizing that intent matters but detection systems can’t always distinguish.
“It's about 40% from an internet page perspective. About a year and a half ago, we looked at Medium and found that over 50% of newly written Medium articles were AI generated.”
“I think the biggest one is norms. It's rude to send undisclosed AI outputs. If someone asks a question and you reply with ChatGPT, it's like you didn’t really engage.”
“Every sentence has dozens or hundreds of ways to phrase it. Over 50 or 100 words, you’re making thousands of decisions. If the vast majority line up with how an LLM writes, it’s vanishingly unlikely a human did it.”
Hosts
Guest
Tracy Alloway
person
Jill Weisenthal
person
Joe Weisenthal
person
Max Spiro
person
Pangram Labs
organization
ChatGPT
product
Claude
product
organization
Quora
organization
organization
Why NASA Hired a Chief Economist
Odd Lots • 49m • 3/31/2026
Javier Blas on Why Oil Could Go Much, Much Higher
Odd Lots • 41m • 4/1/2026
Scott Bok Explains What Investment Bankers Actually Do All Day
Odd Lots • 54m • 4/3/2026
Gina Raimondo on How European Industry Is Getting Crushed
Odd Lots • 45m • 4/6/2026
Search Engine Presents: Are you a good driver?
Odd Lots • 1h 8m • 4/8/2026
Get the full intelligence
Search transcripts, export clips, track mentions, and explore all topics from “This Is How to Tell if Writing Was Made by AI” inside PodZeus.
Start discovering podcast insights today
Start with a 7-day trial and explore a growing catalog of popular podcasts. No credit card required.
No credit card required • 7-day trial • Cancel anytime