Published: 2026-07-22
Substack just released its AI analyzer in partnership with Pangram, and people are already upset because posts they wrote themselves are being flagged as AI-generated.
In reality, some of them used AI and found that the analysis overstated how much was generated by an LLM. And the rest feel like they're being exposed as frauds.
The latter were never good at writing.
AI has exposed bad software engineers. Now it's exposing bad writers.
I'm tempted to call them lazy writers, but bad is more accurate. If they were good writers, they would know what to fix when editing an AI-generated article.
Good writers don't just write. They know how to edit. It doesn't matter whether they're editing their own work, a coworker's work, or a newsletter generated by a computer. Editing is most of the job.
Over time, good writers develop a strong understanding of editing. They drill into every line to make sure the words say what they see in their minds.
Bad writers don't do that. They rely on an LLM to handle the hard part.
AI is great. It takes a huge load of work off our shoulders. We can draft, research, pull quotes, scan articles, and generate posts with a new level of speed. It also puts ideas in front of us that we may not have explored before because the effort was high and the reward seemed low.
I find it funny how upset people are about this. They act like using generative AI somehow invalidates the work.
I write posts in different ways for different newsletters. Sometimes the analyzer says AI was used. Sometimes it says it wasn't.
For my newsletter, I use AI to capture links and transcribe my dictated commentary. It adds a cleaned-up version to a document that I then edit heavily.
I also have a Ghost blog with an automated system that generates each draft. I go through those drafts, edit them, and publish them.
In the first case, I'm spilling my brain out to an LLM and letting it organize my thoughts. In the second, I'm out of the loop until the very end.
The difference isn't whether I'm involved. It's my taste in the final product.
I care about how an article flows, feels, and sounds. Those are things I haven't been able to get an AI to do well. I've tried, but it may never be good enough.
You can prompt all you want, but you won't get out of generative AI what an experienced writer brings to the editing phase. You won't even get out of it what a new writer brings to a first draft.
Yeah, that draft is messy, hard to read, and probably full of cliches. But it has soul. It has that touch of humanity. It has the vibe that's so obviously missing from AI-generated text, even when the human writer is still a novice.
I'm not a professional writer. My bills are paid through software engineering.
That industry has been turned upside down and had the lunch money shaken out of it since the dawn of generative AI. Every few weeks, we hear that AI is going to take our jobs because it can write code.
And, yeah, it does write code. It does it well, too, but only if you know what you're doing. Otherwise, you end up with what's now called "vibe-coded slop."
Software was low-hanging fruit. There's not much human expression involved in determining whether a piece of code is good. Writing is different. Writing is inherently human.
Writing is more than pure logic codified in a mixture of English and special symbols. It's not like a bank password, where any random combination of characters works as long as the system accepts it.
There are an unlimited number of ways to write the same thing, and the reader has to live with the result. If the writing is unclear, the reader pays for it when the 3 a.m. bug takes down the whole system.
Writing has feeling. It has humanity baked into it.
That's why so many people are offended. Some software engineers held back from letting AI generate their code because they felt a loss of control, identity, or pride.
Writing digs deeper. Having AI write your work can feel like admitting you had nothing worth saying in the first place.
That doesn't mean AI should be shunned by the writing community.
If you're reading a blog post and getting nothing from it, move on. If it sounds like AI and you don't like it, move on.
But if you're learning something or gaining a new perspective, the fact that the writer used AI doesn't change the fact that the post helped you.
I still write by hand for the articles on my personal blog. I use AI to generate topics from my slipbox, then riff off those ideas in my own writing.
I might ask AI to check a fact or analyze a piece of writing for ways to tighten it. But the writing process is strictly human because I want the struggle.
Because I want to be a great writer.
It doesn't matter whether the post was fully generated, heavily edited, or written by someone who wrestled with the topic for weeks. If you got something from it, say thank you.