I never publish an AI draft the way it arrives. I also do not rewrite it from the top. I run the same checklist every time, and it does not take long. This post is that checklist, in the order I use it.
First read: does it say anything?
I read the whole draft once without touching it. I ask one question. If a reader finished this post, what would they know or do that they did not before?
If I cannot answer in one sentence, the draft has no point yet. I fix that before anything else. Polishing a post with no point is wasted work.
Cut the warm-up
AI drafts often start with a sentence that could open any article. Something about how fast the world is changing.
I delete it. The post should start at the first sentence that only this post could contain. I do the same at the end. If the last paragraph only repeats the post, it goes.
Check every claim
This is the step I am strictest about. I look for numbers and statistics. I look for names of people, products and studies. I look for statements about what a company or a rule says.
Each one gets checked against a source or removed. A model can state a wrong fact in a very confident voice. My name is on the post, so the mistake would be mine. My longer routine is in how I fact-check an AI draft.
Make it sound like me
I read the draft a second time, slowly, almost out loud. Any line I would not say in conversation gets changed.
I watch for words I never use. I split long sentences. I swap general examples for specific ones from my own work. One real example does more than three paragraphs of explanation.
Fix the frame around the post
A post is more than its body. Before I approve a draft I check four fields:
- The title says what the post delivers.
- The meta description would make me click.
- The slug is short and readable.
- The links point to posts that actually help the reader.
A tool can suggest all four. I still read them, because they are what a stranger sees first. I go deeper on the first two in how I write titles and meta descriptions.
Know when to reject
Some drafts are not worth saving. The topic was weak, or the angle is one I have already covered.
I reject those without guilt. A rejected draft costs me a minute. A weak post on my site costs me trust. In Sastras that is one click.
What I would do first
Write your own checklist of five lines and keep it next to your editor. Use it on the next draft, in order, without skipping. A fixed routine is faster than deciding from scratch what to check each time.