Why Your AI Content Looks Obviously Fake: Try Replacing the Company Name With a Competitor's
Greg Isenberg and Nicolas Cole talked for an hour about AI writing. Cole has made over $30 million writing on the internet, and he says most AI slop is commodity content—you can't tell who it belongs to. The piece includes the full video and a few checks you can use right away.
The term 'AI flavor' has been ruined. Most people use it to mean sentence-level tone, but the real problem is content ownership.
Greg Isenberg posted a tweet with Nicolas Cole's one-hour masterclass on how to write with AI. The tweet says it plainly: the author has made over $30 million from writing online. Greg says this episode is like a free masterclass. Writing courses sound like a wasteland of empty inspiration. This one at least gives you a model you can use to check a draft.
The full conversation is on Greg's YouTube channel, titled 《$30M Writer: Never write AI slop again》. Cole has written nonfiction and ghostwritten for 15 years and runs the SaaS platform Typeshare. In the episode, he gives a framework for judging the value of content and explains why AI will make a lot of writing worthless.
Cole's first observation is counterintuitive: language itself is open source. Letters, words, and names mostly don't belong to any one person. You're named Greg, and he's also named Greg—that's not plagiarism. You write 'the key to losing weight is eating healthy,' and that sentence is fine too, but it belongs to commodity content. Anyone can say it, and AI is best at producing it at scale.
He divides content into three layers: commodity, personality, and original. Commodity content is correct but unowned. Personality content comes from 'I did this,' 'I think this,' and 'I saw this.' Original content is a new idea no one else has said. AI slop feels off-putting often because it stops at the first layer. It reads like an unsigned instruction manual.
Here's a simple dividing line: write 'you should do this,' and the content easily slides toward commodity. Write 'I did this, and here's what happened,' and readers start to see a person. Greg also mentions in the episode that some posts on X get thousands of likes, but when you read them, they're Wikipedia summaries, or an old opinion expanded by ChatGPT. The numbers may be good, but it's playing a different game.
Cole calls point of view a moat. SaaS features are commodity; methodology and point of view are the moat. Typeshare also does writing and scheduling, but Nicolas set a more specific goal for it: let users build a personal language model based on a methodology. When others copy the features, they first have to understand the methodology. Understanding itself takes time.
Some people online question whether AI can really capture voice and flow. The answer in the material doesn't rely on the model improvising. Cole mentions a digital brain, meaning your own language library: stories, opinions, lessons, and common expressions. AI repeats and expands on this approved language, and only then does it sound like you. Let the model read your samples first, then let it write. Give it only one prompt, and it can only guess.
A few checks can be used directly. One person online says: swap the company name for a competitor's name, and if the post still holds up, it's probably slop. Another says: delete the last sentence; the last sentence is often just explaining what the first sentence already said. Reading it aloud also helps—delete any sentence you wouldn't say yourself. Someone online calls this 'human review.' Skip this step, and slop will leak through.
AI has driven the cost of writing close to zero, and the first thing it exposes is who never had a point of view to begin with. Prompts can be adjusted, tone can be changed, but in the end you still have to answer one question: why should this sentence come from you? Don't ask about prompts first. First find one sentence that only you could write.
发布时间: 2026-11-04 22:26