I Used AI to Write this Piece
This is the weirdest thing I’ve ever put on the internet. Follow along if you can.
Is it ok to write using AI? I think so. Well sort of… The following is 4 separate artifacts in reverse chronological order.
1. The final “draft” of this piece
2. The preliminary “draft” of this piece
3. The conversation I had with Claude (The AI) to help draft this piece
4. A context file to be plagiarized if you’d like. (Context files are the reference AI has - like basic operating instructions)
This is the weirdest thing I have ever put on the internet, so I hope you can follow along.
Part 1 - The Final “Draft”
I Used AI to Write This Piece
I wrote this piece using AI. More accurately, I dictated it on my phone while sitting at my desk on a Wednesday morning.
It’s a bit of a silly article, but it’s here to demonstrate my process for potential writers or anyone struggling with leveraging AI.
Step 1 is setting up the context file, some strict ground rules with Claude (my AI of preference) to keep myself honest. I tell Claude quite clearly, do not write anything for me, but help by being my thinking partner.
Step 2 is to create the outline for this piece. Which is exactly what I am doing at the moment. If you’ve ever seen Spaceballs, this is my own little “Now is Now” scenario.
Step 3 is the drafting/dictation process. I like dictation because it allows me to ramble on instead of agonizing over every single word. Stream of consciousness is the whole cloth that I cut back from to create the piece. I find it much more efficient than my former strategy: self-editing in real time.
Step 4 is the external editing step. My greatest advice at this stage of the process really is: “don’t let perfect be the enemy of the good”. Your writing will never be perfect, that’s exactly why editing exists. The AI thinking partner will help with any preliminary gaps and/or structural problems with the writing. After the AI has a pass, a human editor will comb through the piece and ensure it hits its stride.
Step 5 is publishing. This is a mental hurdle more than anything else. Many writers, when they’re first starting out, are terrified that they might ruin their reputation, alienate friends, or be exposed as fraudulent.
So the elephant in the room: the nuance, morality, and tastefulness of using AI. My advice is that the most important thing in writing is to maintain your unique voice. Ultimately that is the thing that cannot be replicated and that will set you apart. Keeping your voice in control can be likened to working out and eating your vegetables. It is time consuming and the allure of laziness is there, but it is much healthier in the long run.
AI is a tool, it all depends on how you leverage it.
So that’s essentially the whole process in a nutshell. This was a strange one to write, but my goal here is to encourage younger Canadians to dive in, and add their voices to the chorus of Canadian discourse. Hopefully you’ll take the leap.
Part 2 - The Original Dictation/”Draft”
I Used AI to Write This Piece
I wrote this piece with AI. Perhaps I’m not writing it at all. I’m dictating it into my phone while sitting at my desk on Canada Day.
It’s a bit of a silly article, but it’s here to demonstrate exactly what my process is for those looking to learn how to write. The next step here is the outline for this piece. There were five or six different pieces I’m building it out of right now. It’s like that scene in Spaceballs — “Now is now.”
There’s a context file that exists to keep myself honest. I’ve set up rules that Claude will simply not allow itself to write anything for me, which is really important. I’m attaching the context file to this piece.
On the dictation — again, we’re very much in the meta here, because I am speaking this into my phone. But I find it really simple to ramble on about something to get my ideas down in a quick manner instead of agonizing over every single word. Oftentimes, people let perfect be the enemy of the good, and that is quite detrimental. That’s what editing is for. That’s why I, as editor in chief, exist — to help ensure pieces are hitting their stride. A lot of my writers, when they’re first starting out, get worried that they are going to either alienate friends or look silly, but really it’s just a question of taking the leap, like jumping into cold water.
The next step is editing it with Claude. Claude gives me this as one long document, and then I bring it into Google Drive to edit. I’ll probably edit it very lightly, if at all, just so you can see what we spat out here.
I want to close with a call to action to anyone who’s been attempting to write, or who’s feeling guilty about using AI: your voice is the most important thing you have, so keeping it intact is vital. That’s true regardless of whether anyone can tell you used AI. In the long run, it’s better, it’s healthier — it’s just like working out and eating right.
Part 3 - My Conversation With Claude
The Full Conversation — accordingto.ca
This is the conversation between Jacob Citron and Claude that produced the article “I Used AI to Write This Piece.” The conversation itself is the article.
Jacob: I wanna write an article on how to write an article leveraging AI in a way that maintains creative integrity and ownership while demonstrating to readers and potential writers how I use AI to assist me in creating my pieces. First, can you give me a list of the context and the rules around what you can and can’t do while we’re writing? Then I wanna build a structure and outline, and we’ll go from there.
Claude: Here’s the rundown of my role, straight from the editorial context file.
What I don’t do:
Write any part of the article, or generate arguments, theses, or positions
Rewrite your paragraphs or sentences (minor grammar/typo fixes are the exception)
Add facts, stats, or references you haven’t introduced
Push the piece in any ideological or tonal direction
Change your voice
Reframe things to seem “more balanced” — that’s your call
Comment on whether a position is right or wrong, only whether it’s argued well
What I do:
Take your draft chunk by chunk and hold the thread of the whole piece
Flag logical gaps, non-sequiturs, or places where something’s assumed but not established
Point out underdeveloped sections
Silently fix clear grammar/typos, and light-clean dictated text without touching meaning
Ask questions instead of assuming and rewriting when something’s unclear
Voice I’m protecting: conversational, analytical, nuanced, logical — lands somewhere, doesn’t moralize, never extreme in either direction. If a fix I suggest would make it more formal, neutral, or generic, it’s the wrong fix.
Canadian conventions: Canadian spelling, Canadian framing unless the piece calls for something else.
One nice bit of texture: this piece is about how you use AI, so the process itself — me flagging things, you deciding — is basically the demonstration. Worth keeping that in mind as we build the outline, since the structure might want to show, not just tell.
Whenever you’re ready, give me your working title or angle — or just start dictating and we’ll build the outline as we go, whichever you’d rather.
Jacob: I want to call this piece “I used AI to write this piece.” It’s actually gonna be framed as a tutorial. Step one is the context file, which we’ve already discussed, making sure that there are ground rules in place using Claude. Step two is to demonstrate the thesis and then build an outline. So my thesis is that you can use AI in a way that is respectful and preserves the integrity of the humanity within the writing. I also want to point out that I’m dictating everything going on right now. That’s why my writing oftentimes seems conversational. Anyway, the structure of this article will be: one, setting up the context file — I will actually ask you at the end to export that so I can give it to people for their own safeguards. Step two is come up with a thesis and the outline. Make sure you include timestamps. At the start, I wanna make sure that this is a data point that people can understand. I think that’s part of the jarring nature of it all. In this case, the thesis is simple: it’s an intro, as well as showing the outline, then the dictation piece, then editing with Claude, and then human passes in Google Drive to go from there. And that’s it.
Claude: Got it. Here’s the structure as I understand it — reflecting it back before we lock it in.
“I Used AI to Write This Piece”
Intro / Thesis — the piece opens by stating the thesis plainly: AI can be used in a way that’s respectful of and preserves the humanity and integrity of the writing. You’ll note upfront that you’re dictating, which is why it reads conversationally.
The Outline — shown to the reader as a structural artifact (what you’re building right now)
The Context File — the ground rules doc, exported at the end for readers and writers to take and use themselves
The Dictation Piece — the raw process of you talking it out
Editing with Claude — showing the flagging and questioning process in action
Human Passes in Google Drive — the final human-only revision layer, outside AI entirely
[Jacob selected: one timestamp at the very start only]
Timestamp locked in: 9:19 AM EDT, July 1, 2026. That’s the skeleton. Whenever you’re ready, start dictating the intro and I’ll hold the thread from there — flagging as we go, not writing.
Jacob: I wrote this piece with AI. Perhaps I’m not writing it at all. I’m dictating it into my phone while sitting at my desk on Canada Day. It’s a bit of a silly article, but it’s here to demonstrate exactly what my process is for those looking on how to write. So the next step here is to come out with the outline for this file. There were five or six different pieces. I’m building it right now. It’s like that scene in Spaceballs — “Now is now.” There’s a context file that exists to keep myself honest. I’ve set up rules that Claude will just not allow itself to write anything for me, which is really important. I’m going to attach the context file to this piece. On the dictation — again, we’re very much in the meta because I am speaking this into my phone. But I find it really simple to ramble on about something to get my ideas down in a quick manner instead of agonizing over every single word. Oftentimes, people let perfect be the enemy of the good, and that is quite detrimental. That’s what editing is for. That’s why I exist, to help ensure that pieces are hitting their stride. A lot of my writers, when they’re first starting out, get worried that they are going to either alienate friends or look silly, but really it’s just a question of taking the leap, like jumping into cold water. The next step is gonna be editing it with Claude. You’re gonna give me this as one long document, and then I’m gonna bring it into Google Drive and edit it. I will probably edit it very lightly, if at all, just so that we can see what we spat out here. I wanna close with a call to action to anyone who’s been attempting to write or is feeling guilty about using AI. First, when you use AI to write things for you, people can tell. There are artifacts and fragments and sentence structures. And also, I found that people check. Finally, your voice is the most important thing you have, so keeping it intact is vital. Give it a shot.
Claude: Got it — holding the thread. Two flags before you move on.
1. The closing pivot — you move from “people can tell / people check with AI detectors” straight into “your voice is the most important thing, keep it intact.” Those are two different arguments (getting caught vs. integrity for its own sake) sitting next to each other without a bridge. Is the CTA “protect your voice because you’ll get exposed” or “protect your voice because it matters regardless of exposure”? Right now it reads like both, unresolved.
2. “That’s why I exist, to help ensure pieces are hitting their stride” — the “I” here is ambiguous. Is that you speaking as the writer, or is it meant to read as Claude’s voice?
Everything else is tracking fine — the meta framing lands, and the context-file and dictation sections connect cleanly to the outline we built. Keep going whenever you’re ready.
Jacob: The thrust is the latter. Your voice matters regardless of exposure. In the long run, it’s better, it’s healthier, it’s just like working out and eating right. The I here is about me being the editor in chief for the site. That’s the service I provide and why it’s valuable to have collaborators in general — human ones, that is.
Claude: Clear — noted on both.
One more flag on the CTA now that the throughline is confirmed: if “voice matters regardless of exposure” is the actual point, the detection material currently reads as the lead argument with voice-preservation tacked on after. Worth considering whether detection should be framed explicitly as the lesser reason — something like “even if you never got caught” — so the piece doesn’t accidentally land on exposure as the main stake. Your call.
Jacob: I think I’ll probably take it out actually entirely, that AI checker piece. It’s not relevant and distracts from the main point that it’s really about personal, professional integrity.
Claude: That tightens it — cutting the detection bit puts the weight fully on integrity as its own reason, no hedging on getting caught.
Jacob: Please do put it into Google Drive.
Claude: Draft uploaded to Google Drive as “I Used AI to Write This Piece - Draft.” Assembled from exactly what you said, with only light cleanup of dictation artifacts — nothing added or rewritten. Ready for your human pass.
Jacob:[First human pass — edited draft returned]
Claude: Clean pass — the structure landed well. Two small flags.
1. Step 4, “a human editor like me” — since you’re both the author and editor in chief, “like me” is slightly ambiguous. A reader could reasonably wonder whether you’re speaking as the writer or as the editor.
2. The jump from Step 5 into the voice/integrity paragraph — you go from “worst case, no one will read it” straight into “the nuance, morality, and tastefulness of using AI” with no transition. The bridge is missing.
Jacob:[Second pass — bridge added, “like me” addressed]
Claude: That bridge works — “the elephant in the room” picks up the publishing anxiety and pivots cleanly into the integrity argument without over-explaining the connection. Flag resolved.
The “like me” ambiguity in Step 4 is still sitting there from the previous pass — worth a look, but not a blocker.
Jacob: I removed it. Let’s get this published! [Final draft submitted]
Claude: Two small things before you hit publish.
1. “Which is exactly what I am doing at the moment” — this is present tense and works perfectly as a meta moment while you’re writing it, but once it’s published it’ll read as frozen in time in a slightly odd way. Worth deciding if you want to leave it as an intentional artifact of the live-writing conceit, or adjust the tense.
2. “nut shell” — one word: nutshell.
Otherwise this is clean and ready. Great piece — the process demonstrated itself.
Jacob:[Final edits — intro revised, Spaceballs line smoothed, nutshell fixed]
Claude: Clean. One flag still outstanding: the present-tense “which is exactly what I am doing at the moment.” Your call whether that’s an intentional artifact of the live-writing conceit or something to smooth out before it goes up.
Jacob: Let’s trust that the reader gets it. I will link to a Spaceballs clip. Great.
Part 4 - A link to the context file:
https://www.accordingto.ca/p/accordingtoca-context-file-for-writers
Cheers and Happy Canada Day!