Ha! Mark, this is more personal than universal. Some people use 'leverage' and 'synergy' unironically and sleep just fine at night. (I don't trust those people, but they exist.)
My blacklist is built from words that make MY writing sound like someone else wrote it. But here are a few that trip up almost everyone because AI loves them way too much:
Delve
Utilize (just say 'use,' we're not filing a patent)
Landscape (as in 'the AI landscape')
Robust
Seamlessly
Leverage (as a verb)
Navigate (when nobody's on a boat)
Cutting-edge
Elevate
Foster
The real move isn't copying my list though. It's reading your last five AI drafts and circling every word that made you wince. That's YOUR blacklist. The one that matters.
The VAST framework is genuinely brilliant for this problem. That observation about the difference between 'vibes' and 'specs' cuts right to why most writers feel like they're slowly losing something they cant quite name. In my own experience trying to get AI to match certain tones, I kept saying 'make it punchier' and wondering why different models interpreted that wildly differently. Documenting the actual patterns instead of the feeling is such a obvous solution in retrospect.
Have you found any specific words that always seem to confuse the AI? Also I have tried to dumb the AI down to having it list things that a 5th Grade US student could read and then use thoose points to add in "what sounds like me". It sometimes streamlines the process so I can jump in witth quick lines that I would use to explain things.
The 5th grade readability trick is actually quite smart. You're forcing the AI to strip out its default corporate voice before you layer yours on top. Clean canvas first, then your fingerprints.
The words that confuse AI the most aren't complex ones though. It's the vague ones. "Conversational" means something different to every model. "Punchy" is a coin flip. "Authentic" is basically a Rorschach test.
What works better is constraints instead of descriptions. Not "make it conversational" but "sentences under 12 words, no semicolons, start two paragraphs with And or But." The AI can't misinterpret a rule. It can absolutely misinterpret a vibe.
Sounds like you're already halfway there with your process. The next step is writing down those "quick lines that sound like you" as a permanent reference doc instead of re-explaining them every session.
Claude and I have a little ongoing tussle over while vs. whilst. I make it use British spelling conventions and it insists that "whilst" is correct, but ugh, not in my writing it isn't. In every single draft, it changes the word, I change it back, it changes it again. I should really adjust Skills, change project instructions blah blah, but it's become a little ritual, lol. Sometimes it slips through, it's not wrong but it is Claude winning that round 😂
P.S. Shocked you would bet the dog but guessing you are very very sure! 🫨😂
Nick! My Gee! Do you have specific "blacklist" of words most people should start with?
Ha! Mark, this is more personal than universal. Some people use 'leverage' and 'synergy' unironically and sleep just fine at night. (I don't trust those people, but they exist.)
My blacklist is built from words that make MY writing sound like someone else wrote it. But here are a few that trip up almost everyone because AI loves them way too much:
Delve
Utilize (just say 'use,' we're not filing a patent)
Landscape (as in 'the AI landscape')
Robust
Seamlessly
Leverage (as a verb)
Navigate (when nobody's on a boat)
Cutting-edge
Elevate
Foster
The real move isn't copying my list though. It's reading your last five AI drafts and circling every word that made you wince. That's YOUR blacklist. The one that matters.
The VAST framework is genuinely brilliant for this problem. That observation about the difference between 'vibes' and 'specs' cuts right to why most writers feel like they're slowly losing something they cant quite name. In my own experience trying to get AI to match certain tones, I kept saying 'make it punchier' and wondering why different models interpreted that wildly differently. Documenting the actual patterns instead of the feeling is such a obvous solution in retrospect.
Curious: when you were saying “make it punchier,” did you eventually figure out what you actually meant mechanically?
That translation moment, from “I know it when I see it” to '“here's specifically what I mean,” is where most people get stuck.
And honestly where the most useful patterns are hiding.
Have you found any specific words that always seem to confuse the AI? Also I have tried to dumb the AI down to having it list things that a 5th Grade US student could read and then use thoose points to add in "what sounds like me". It sometimes streamlines the process so I can jump in witth quick lines that I would use to explain things.
The 5th grade readability trick is actually quite smart. You're forcing the AI to strip out its default corporate voice before you layer yours on top. Clean canvas first, then your fingerprints.
The words that confuse AI the most aren't complex ones though. It's the vague ones. "Conversational" means something different to every model. "Punchy" is a coin flip. "Authentic" is basically a Rorschach test.
What works better is constraints instead of descriptions. Not "make it conversational" but "sentences under 12 words, no semicolons, start two paragraphs with And or But." The AI can't misinterpret a rule. It can absolutely misinterpret a vibe.
Sounds like you're already halfway there with your process. The next step is writing down those "quick lines that sound like you" as a permanent reference doc instead of re-explaining them every session.
That way the spec survives the next model swap.
Claude and I have a little ongoing tussle over while vs. whilst. I make it use British spelling conventions and it insists that "whilst" is correct, but ugh, not in my writing it isn't. In every single draft, it changes the word, I change it back, it changes it again. I should really adjust Skills, change project instructions blah blah, but it's become a little ritual, lol. Sometimes it slips through, it's not wrong but it is Claude winning that round 😂
P.S. Shocked you would bet the dog but guessing you are very very sure! 🫨😂
Claude playing the long game on "whilst" is honestly impressive. It knows you'll get tired before it does.
(And Butters remains safely un-wagered. He's my ace in the hole for when I need to sound confident about something I'm only 97% sure of.)