There's a question people are asking about your content whether they say it out loud or not: how much of this did AI write?
I know they're asking because I ask it too. I'll read a post, hit a paragraph that's a little too tidy, and catch myself wondering. It's not always a criticism. It's just a thing my brain does now. And if I'm doing it to other people, I have to assume other people are doing it to me.
The judgment is already happening. The only real question is whether I get a say in it.
The judgment is happening with or without me
When someone can't tell where a piece of writing came from, they don't suspend judgment. They guess. And a guess made in the absence of information tends to land badly. They assume more AI than there was and discount work I actually did, or they assume less and give me credit I didn't earn.
Both outcomes are bad for me. The first one costs me credit, and/or authority. The second one is a debt that comes due the moment someone finds out.
I choose transparency
Good intentions die on friction. If ai disclosure meant writing a fresh little disclaimer every time, I'd do it for a week and then quietly stop.
So I built a page: tregrisby.com/ai-levels. It's a simple 0-1-2 scale with copy-ready labels for social posts and email. I tap the one I need, it's on my clipboard, done.
AI Level 0 — No AI. Written entirely by me.
AI Level 1 — Hybrid. Human-led, AI-assisted. I wrote it; AI helped polish. There's also an accessibility variant for when AI was only used for formatting and accessibility support.
AI Level 2 — All AI. Fully AI-crafted from my prompts, with a variant that notes I reviewed it before sending.
Why disclose at all
Because I want the trust, and trust is upstream of everything else.
If you know how a piece was made, you can decide whether you agree with it on its merits. That's what I want. Read the content, weigh it, tell me what you think. What I don't want is someone forming an opinion about my work based on a story they made up.
Disclosure isn't apologizing for using AI. It's removing a variable so the actual content can be evaluated.
Not everything gets the same treatment
There isn't one correct standard for all content. Some material needs to be mine. If it's a technical recommendation, a claim about my own work, an opinion I'm going to be held to then it gets my manual input on it, start to finish. The value of it is that I built it and reviewed it. Level 0 or a very deliberate Level 1.
Other material is either high-volume, low-stakes, or both. Reformatting notes. A calendar description. The fourth variation of a caption. "Close enough" is a completely legitimate standard for that work, and pretending otherwise means burning judgment I need elsewhere. That's Level 2, and I'm not embarrassed about it.
The skill isn't "use AI" or "don't use AI." It's knowing which bucket a piece of work falls into.
This is a transition, not a permanent state
I don't think I'll be doing this forever.
I remember when "I read it on the internet" was a punchline. The source itself was disqualifying. That skepticism was reasonable at the time and it faded over a couple of decades. The internet didn't get more trustworthy, but people got better at evaluating what was actually in front of them. The question shifted from where did this come from to is this any good?
I think AI goes the same way. Eventually the how-much-was-AI question gets boring, and what's left is the only question that ever mattered: did "it" work? Did "the message" land? Did the reader get what they came for?
If that takes a little AI, fine. If it takes none, fine. If it takes maximum Ai, that is also fine. Delivering something is important. "Perfect is the thief of good."
The part that doesn't get to fade
The thing I'd want someone to take from this post:
I don't get to blame the AI.
If my name is on it, it's mine. Every claim, every mistake, every clunky sentence. "AI wrote that part" is not a defense — it's an admission I shipped something I hadn't read carefully enough. A Level 2 label discloses how a thing was made. It does not transfer responsibility for it.
The stamp is the promise. I approved this, flaws and all.
That's why the labels don't make me nervous. They're a note about process from someone who's standing behind the output.
Steal the Ai scale levels
The page is public and the labels are copy-ready: tregrisby.com/ai-levels. Take them, change them, make your own version with different names.
And if you're building something and want a developer who's this deliberate about the tools — that's what I do.
⚖️ AI Level 1 · Human-led, AI-assisted
(Adjust to whatever's true after your edit pass, and put it on the post.)
Top comments (0)