May 31, 2026
Bot, Actually
It's Not Just X. It's Y
Readers say AI writing has a tell — and now everyone’s side-eyeing their own sentences
TLDR: The article argues a common phrase isn’t bad writing by itself, but panic over AI-generated text is pushing people to rewrite their own words just to seem human. In the comments, some call these phrases a dead giveaway, while others cheer them on as handy “watermarks” for catching bots.
The article starts with a deceptively tiny phrase — “It’s not X, it’s Y” — and turns it into a full-blown culture war over what “real” writing even looks like anymore. The author argues this sentence style isn’t automatically bad; it’s a classic rhetorical trick humans have used forever. The real scandal, they say, is that people are now so paranoid about machine-made text that perfectly normal writing is getting treated like contraband. Even grammar tools are reportedly nudging writers to rewrite harmless phrases just to avoid being flagged by AI detectors, which the author describes less as helpful editing and more as paying protection money for your reputation.
And oh, the comments came in hot. One camp basically yelled, “This is the smoking gun!” after a commenter suggested the phrase might be spreading because of how these systems are trained. Another crowd has gone full amateur detective, listing supposed AI “tells” like “No X, No Y, No Z” slogans, over-polished README lines, and the dreaded flood of ✅ and ❌ emojis. But not everyone hates the robot tics: one commenter actually said they like these weird phrases because they act like watermarks, making machine text easier to spot. Then came the funniest dunk of all — the reminder that early “reasoning” in AI often looked suspiciously like a bot muttering “Wait…” and “Hmm…” to itself, as if adding fake inner monologue could pass for deep thought. In other words: the internet is now auditing vibes, one phrase at a time.
Key Points
- •The article identifies “it’s not X, it’s Y” as a negative parallelism structure that large language models often generate.
- •It argues that this rhetorical device is not inherently bad writing and cites JFK as an example of its established use.
- •The article says Grammarly flagged multiple phrases in the text as likely AI-generated and suggested alternate wording.
- •It states that the author paid Pangram to check that a journal article would not be flagged as AI-generated before submission.
- •The article links recurring LLM language patterns to post-training methods, specifically RLHF and RLVR, rather than to raw web training data alone.