How do you keep a data blog from sounding like every other AI-written post?

I hit exactly this wall. The fix that worked: stop letting the data lead. Generic data writing opens with the dataset and walks through what it shows — and that's precisely the shape AI produces by default, which is why it reads as generic. What I do now is open with something only I have — a place I actually spent time in, a project I actually worked on, a call I actually got wrong — and let the data come in to support or complicate that. The numbers serve the story instead of being the story. AI can analyze any dataset competently; it cannot have been somewhere or made a specific mistake. So the distinctiveness isn't in the analysis — it's in the lived part the analysis is attached to. Lead with that, and the same data suddenly reads as yours.