Do I tell people AI helped write this?
I get this question more than almost any other, usually from someone who just found out a blog post or a script they liked had an AI model somewhere in the pipeline. The tone is rarely hostile. It's more like they want to know if they got tricked. Fair question. Here's my honest answer, and the reasoning behind it, not a PR line.
The short version
Yes, I disclose AI involvement when AI did meaningful creative or factual work. No, I don't slap a disclaimer on every sentence a model touched, because if I did that, half the internet would need one on the punctuation.
The real issue isn't whether AI was "involved." Almost every piece of writing I publish has touched AI somewhere, whether that's a grammar pass, a research summary, or a full first draft. The issue is what role it played and whether hiding that role would change how you read what I wrote.
What "AI helped write this" actually covers
This phrase gets used for wildly different things, and that's where a lot of the confusion comes from.
At one end, someone types a topic into a chatbot, copies the output, and publishes it with maybe a light edit. At the other end, someone writes every sentence themselves and just runs it through a model to catch typos, the same way people used to run things through Word's spell check. Both technically involve AI. They are not the same act, and they don't deserve the same disclosure.
In my own workflow, I run a mix of local and cloud models depending on the job. Local models handle things like transcript cleanup or repetitive formatting where I don't need frontier reasoning and I'd rather not send that data out. Cloud models get used for research synthesis, drafting, and structuring an argument when I need more capability than what I can run on my own hardware. I also run agents that chain steps together, pull from source material, and hand off a draft for me to edit. None of that is a secret. But none of it means the article "wasn't written by me" either, since I'm the one deciding what goes in, what gets cut, and whether the claim is actually true.
Why a blanket disclaimer doesn't solve anything
A lot of sites have started adding a boilerplate line like "this article was created with the assistance of AI" at the top of everything, regardless of how much or how little AI actually did. I don't do that, for a practical reason: it conveys almost no information.
If you tell me every article on a site has that label, I learn nothing about any specific article. I don't know if it means a model wrote the whole thing unsupervised, or if it means someone used autocomplete. A disclosure that applies uniformly to everything stops functioning as a disclosure. It becomes a legal hedge, not information for the reader.
What actually matters to a reader is narrower: did a human check the facts, does the opinion in this piece reflect a real person's view, and was anything published without a human reading it first. Those are the questions worth answering, and a generic banner doesn't answer any of them.
Where I draw the line
I disclose specifically, not generically, and I draw the line based on origin of judgment, not origin of words.
If an article states a claim about how a tool works, what it costs, or what results it produced, that claim came from me checking it, not from a model asserting it with confidence. Models are fluent, not accurate. They will generate a specific-sounding number or a benchmark that feels real, and it's my job to strip that out if I can't verify it myself. This is the actual production discipline behind everything I publish. If I haven't verified a number, it doesn't go in, whether a human or a model typed it first.
If a video script includes my opinion on a workflow or a piece of software, that opinion is mine. A model can help me phrase it, help me structure the argument, or draft a version I then rewrite, but the position itself isn't outsourced. I wouldn't publish "I think this tool is worth using" if I hadn't actually formed that view by using it.
Where I do flag things explicitly is when a full piece was substantially AI-drafted with light human editing, because in that case the origin of the words is different enough from my normal process that I think it changes how you should weigh what you're reading. That's rare for long-form articles here, more common for short repetitive content like listing formats or routine updates, and I'd rather say so than let you assume every piece went through the same amount of hands-on work.
Why this connects to how AI actually works, not how it's marketed
A lot of the disclosure debate gets tangled up in bigger claims about AI that I think are overstated. I'm not going to tell you AI is going to replace writers, and I'm not going to tell you it's some inevitable superintelligence that makes human judgment obsolete. What I run day to day is a set of tools that are very good at pattern completion and structuring language, and very unreliable at telling the difference between something true and something that merely sounds true.
That gap, between fluent and accurate, is exactly why disclosure matters more for factual claims than for style. A model helping me phrase a sentence more clearly poses basically zero risk to you as a reader. A model inventing a statistic that slides past my editing because it sounds plausible is a real risk, and that's the failure mode I actively guard against, not something I gloss over.
This is also why I don't think "AI wrote it" is even the most useful framing. The more useful question is "did anyone check it," and that's true whether the first draft came from a model, from me typing fast, or from a WordPress import I didn't review closely enough. AI just makes the volume of unchecked content easier to produce at scale, which raises the cost of skipping the check, not the cost of using the tool.
What this looks like in practice on this site
Practically, here's what you can assume unless I say otherwise. Claims about how something works, what a tool does, or what I've observed running it are things I've verified myself, regardless of what tooling helped me write the sentence. Opinions are mine. Full AI-drafted pieces with minimal editing get a note saying so. Local versus cloud AI comparisons reflect tools I've actually run, not summaries pulled from other people's reviews.
I'm not asking you to trust me because I say "trust me." I'm telling you what the actual editorial process is, so you can decide how much weight to put on any given piece. That's a more honest answer than a checkbox that says "AI-assisted," because that checkbox would be true for almost everything and would tell you almost nothing.
If you want to see the actual tools and workflows behind this, not just the philosophy of it, that's what the rest of the site is for.
Head back to the [homepage](/) to see how the automation, agents, and local versus cloud AI setups actually run.
Get new guides and videos first — join the Telegram channel.