I keep thinking about the old secretarial disclaimer: “Dictated but not read.”

A small sentence with a big message on status. Someone important had spoken. Someone else had typed. If the result was awkward, inaccurate, or barely coherent, the reader had been warned.

It suggested the person who dictated was in the Alpha position.

Some AI output (perhaps slop) is the modern version, except we have removed the typist who might have fixed the worst parts.

The process is simple: type a prompt, copy the answer, publish it under your own name. The grammar is clean. The headings are orderly. The author may not believe any of it.

That’s the problem.

I use AI every day. It is great for getting past the blank page, organizing a pile of notes, and showing me where an argument is thin.

AI lowers the cost of producing a draft. It does not remove the need for judgment.

Too much of what we are publishing now has no judgment in it at all.

Content as administrative output

Most AI output reads less like writing and more like paperwork. It exists because the content calendar required a post, the SEO plan required another page, or someone decided the company needed to “show thought leadership.”

So the machine fills the space.

It produces 1,500 words where 300 would have done. It repeats the accepted wisdom, rounds off every sharp edge, and concludes with a cheerful paragraph about the future. Nobody is offended. Nobody is challenged. Nobody remembers it five minutes later.

The dashboard still records a publication.

That’s how writing becomes bureaucracy. The purpose stops being communication. The purpose becomes proof that content was produced.

The automation of indifference

The problem with AI slop isn’t that a machine helped write it. The problem is that no human felt responsible for the finished piece.

Nobody checked the claim.

Nobody asked whether the argument was useful.

Nobody noticed that the conclusion contradicted the introduction.

Nobody thought about what should be left unsaid.

We save a few minutes by skipping that work, then make every reader spend time deciding whether anything on the page can be trusted. That’s not a good trade.

And it gets worse as these systems begin reading and summarizing one another. One machine produces the article. Another condenses it. A third cites the condensation. By the time a person sees the answer, the original thought may never have existed.

Just like the infinite monkey theory, there’s no Shakespeare here.

Read what carries your name

AI can draft. It can reorganize. It can help find a better way into an idea.

The person whose name appears at the top still owns every sentence.

Read it. Question it. Cut the parts you do not mean.

If you cannot be bothered to read the writing that carries your name, the rest of us probably should not bother reading it.