On Ai Writing

21 July 2026 by Phillip Johnston

One of the more maddening parts about the widespread adoption of AI tools, for me, is the inescapability of AI writing. I see it in comments, posts, and newsletters. I even see my colleagues using it in their published articles.

I agree with Zvi Mowshowitz on the problems with AI writing:

At this point, I see AI writing as having four kinds of issues.

1. The writing is objectively bad in various ways, especially low perplexity and low density of information.
2. The writing is all the same style, with diminishing marginal returns.
3. The writing gives away that it is what it is, in ways you cannot unsee, and no one does the work to try and hide this.
4. If you don’t write, you’re not thinking and learning.

Sure, it’s “productive,” and perhaps it gets content out that wouldn’t otherwise exist. Maybe you even have novel ideas reflected in the AI-generated text. But let’s think about the tradeoffs.

Once detected, I will stop reading whatever is in front of me. But even more, I will stop reading you forever. This is a catastrophic blunder if you’re trying to build a reputation, a brand, or a business.

It also destroys your differentiation. I want to hear your voice, I want to receive your insights and experiences, painfully and surprisingly and joyfully earned. If you publish something under your name and headshot that sounds just like everyone else’s AI writing, well, it’s hard for me to view you as anything other than yet another bot to be ignored.

For our own field, I have to wonder too: what does it do to a technical community when experts outsource their thinking to AI tools? My experience agrees with Zvi’s take, which is that writing is a great driver of refining our thinking and understanding. There is value in the hard work of wrestling with the ideas, finding ways to communicate them effectively, and identifying holes in the thinking, and that work happens when we sit down to write. The hard-won judgment, shared effectively, is the whole point – not publishing articles just for the sake of publishing them. If an article is AI-generated, how do I even know that you understand what you’ve published?

I believe there are ways to use AI to strengthen your writing, especially by stress-testing your ideas before you share them. I don’t think using tools to improve your work is the problem. The problem is publishing unedited, un-thought-through output and claiming it as your own voice. Don’t do that. I want to read your words.

The post On AI Writing appeared first on Embedded Artistry.