Writing
Is It AI Slop? The Truth About How I Write
"AI slop" is a useful phrase because it names a real failure mode: text that sounds polished but has no lived thought behind it. The mistake is assuming every AI-assisted piece is automatically slop. The better question is where the judgment came from.
My writing process is closer to steering than outsourcing. I use AI to shape drafts, pressure-test structure, and surface alternate phrasings. The actual authorship lives in the selection: what I keep, what I delete, what I sharpen, and which distinctions I refuse to flatten.
This matters more now, not less, because AI makes mediocre fluency cheap. That raises the bar for real writing. A piece has to carry a point of view, a lived constraint, or a useful distinction — otherwise it is just competent noise. So judge AI-assisted writing by the specificity of the thinking, not by whether a model touched the prose.
The authorship test
The test I use is whether the piece contains a real selection function. Did a human choose the frame, reject plausible-but-wrong simplifications, preserve the uncomfortable distinctions, and take responsibility for the final claim? If not, the prose may be fluent, but it is unowned.
The selection matters because AI optimizes toward averaged coherence. Good writing often depends on the non-average move: a sharp distinction, a lived constraint, a specific example, or a refusal to round the argument into a smoother but weaker version. Sometimes the rough edge is where the truth lives, and the job is to keep it.
How I actually do it
- Start from a thesis or tension I actually care about.
- Use AI for expansion, counterarguments, structure, and compression — to create contrast, then pick the version with teeth.
- Manually preserve the distinctions the model tries to flatten.
- Cut paragraphs that sound good but do not move the argument.
AI can help with the language. It cannot supply the reason the piece needed to exist. That still has to come from the writer.