Thoughts on Letting AI Write My Posts
My last few posts were written by AI. This one I typed the old-fashioned way.
I had AI auto-generate a few articles and published them on WeChat. How did it go? Some AI pieces got solid traffic; some got almost none. So traffic still comes down to whether the audience cares about the topic — not whether the prose is fancy, and not whether a human or a model wrote it. As long as the idea lands, people stay on the page a little longer.
I tried a few kinds of content:
- Open source (my own projects and other people’s)
- AI
- Infrastructure
I figured AI topics would do better. The opposite happened. AI is a hot keyword, but posts that dig into AI details have a tiny audience — the recommendation algorithm barely pushes them either. Maybe the writing was weak; maybe I picked the wrong angles. Either way, traffic was low. My followers use AI, but they don’t obsess over the internals. Tips on how to use AI, which models feel better day to day — that might work better. My take: applied / tool-style AI content gets more reach; detail and theory pieces don’t, unless you’re already an AI-native brand. That’s vertical traffic for you.
Open source usually doesn’t tank — the IT circle is wide enough. Posting my own projects tends to outperform recommending other people’s, though it still depends on the topic. Call it fifty-fifty.
Infrastructure clearly beats the other two for me. My content and my audience skew toward ops, SRE, and platform engineering. Infra news and tools aren’t always useful to readers, but people want to read them, can follow them, and will talk about them. That’s why the numbers look better.
AI drafts are fast, tidy, and logical — language models are built to predict the next token and keep things coherent.
Humans wander. Unless you force a framework, the mind branches. Non-academic writing ends up with fingerprints: favorite phrases, layout habits, tone. Everyone sounds different.
AI can mimic a human voice — a bit wordy on purpose, the odd grammar slip — and that can fool people for a while. But AI still has tells. Mass-produce enough posts and the pattern shows.
What worries me more: AI is a great writing tool, but if you never judge or think through what it produces, you skip verifying accuracy. For a tech post I still have to check the official docs, click the links, confirm the facts. AI’s logic is also stiffer than a person’s. Even with skills and rules, after a few generations the scaffold, wording, and flow look almost identical — read one paragraph and you can guess what it will write next. And if you stop judging and thinking, your own expression, judgment, and writing will slide.
People are lazy. Laziness is default. Generate it, skip the mental work, think “wow AI is great,” and the information skims the cortex like a breeze over water — a ripple, then flat again. Right?
So what am I trying to say? Lately AI-assisted posts didn’t tank my traffic, which surprised me — so use AI, but don’t hand everything to it. And for people who actually want to share and write, traffic isn’t the main score. Clarity matters more: did you say what you meant, and did it reach the readers you hoped for?
That brings Miyazaki to mind — frame by frame, hand-drawn, years of patience for work he was proud of. Even after digital painting arrived he stuck with hand drawing, though later one film used CG too. Trends are trends: AI helping people hit goals isn’t something you can (or should) fight. You can still choose what you make. The core is the content — what you’re trying to say. Form can change a thousand ways; the core shouldn’t, and the core is what counts.
This post, for example: AI is fine as an assistant to help you say what you mean. Don’t let it replace your thinking. Don’t let it own what you want to express.
That’s my recent take on using AI to write. Yeah. See you next time. And — do you think this one was written by AI?