We Can't Agree on What Counts as "AI-Written"

bookriot.com

7 points·theanonymousone··4 comments

It's going to get nearly impossible to avoid the taint of some AI, at least for a while. The article first mentions Grammarly, a tool that up until was a normal English language grammar and style checker, widely used and appreciated by many writers. In 2025, they added something nobody asked for. Now it's a vector for AI taint, and given how widely it's used, there's untold numbers of tainted documents out there now.

The article doesn't mention, however, the incorporation of AI into the writing tools themselves. Copilot in Word means that everything written with it is suspect. Google has shoved Gemini into everything it can. Do we believe everyone when they say they didn't use AI assistance? Probably not, because even the spell and grammar check is suspect now.

I don't know if its just me, I recall feeling even before LLMs took off that Grammarly was pushing some to edit their writing in ways that just screamed "assisted by Grammarly".

I loathe to contemplate just how rough it is these days, at least basic LLM output still carries some rather obvious signals. Its gotten to the point I use an extension to count many of them and will outright entirely avoid anything that trends too high before I even lay eyes on the text.

"assisted by Grammarly"

Yeah, it doesn’t take AI to recognize throngs like passive voice, excessive wordiness, cliché phrases, and run-on sentences. But the writer was free to reject those suggestions.

“ Look, I get it. It’s Q3 2026, and we should expect that everyone is utilizing AI at SOME point in their process (sourcing ideas, creating outlines, refining prose, etc.).”

Well no, everybody is not idle or incompetent enough to outsource these basic skills to an algorithmic content thief.

Reply on news.ycombinator.com