Does banning "not X, it's Y" actually fix the problem?
AI always looks for a workaround. Say hello to negative alternative logic
I've written a lot about "not X, it's Y" contrastive logic – it's close to the default setting for ChatGPT and Claude. Brands see it, think it sounds smart, and litter their content with it. Multiply that across thousands of brands and you get a marketing ecosystem built on the same tedious logic.
Some brands have noticed this and order their robot to exclude contrastive logic. Plenty of content marketers have spotted it too and will bang on about it like they've reinvented the theory of relativity. The usual fix is prompting AI not to use contrastive framing at all – no "not X," no "isn't just this, it's that."
But eradicating contrastive logic through prompting just tells AI what not to do, without explaining what to do instead. So the robot's basic logical habit morphs into something else rather than disappearing completely. It just finds a workaround.
I call what spawns from this negative alternative logic. It still redefines what's "really" true by proposing a decoy first – it just strips out the words that give contrastive logic away.
Compare these two sentences that my robot wrote for me:
"Patient no-shows aren't a scheduling problem – they're a communication problem."
"Patient no-shows are rarely a scheduling problem; they're usually a communication problem."
Identical rhetorical move: both open with a decoy diagnosis, then redefine what the "real" problem is. It doesn't matter to the AI whether that decoy has any grounding elsewhere in the piece. As long as the sentence sounds like a considered alternative was weighed and ruled out, that's good enough for the robot – even though eradicating the "X" problem was never the point in the first place.
It's like saying in a football match report that:
"Harry Kane didn't score because he second-guessed the defender, he scored because of the defender's error."
"Harry Kane's finish looked like a moment of individual brilliance. In reality, it was the defender's mistake that made it."
Only the first sentence gets caught by anyone checking for contrastive language. Brands miss the second because "rarely," "usually," and "in reality" aren't on their banned-word list. The decoy's still there, doing exactly the same job.
Why am I writing about this? Because I see it happen to me all the time, and I have to keep building extra prompts to stop my robot (or my clients' robots) finding a workaround.
I use AI every day for batch content production, and banning contrastive language is always on my prompt list. But prompting alone isn't enough – the robot always finds a different route to the same lazy logic. That's why AI content editing matters for brands that want a voice readers won't tire of halfway through an article.
Email joe@shortstory.ltd if you want a batch content editorial service that casts a critical eye on AI.