AI Use Policy

I use large language models in the drafting of all articles published on this website. This page explains what that means in practice, and why I disclose it this way.

What I do, and what the AI does

I set the scope of every piece: the question worth asking, the argument worth making, the evidence worth weighing. The model drafts and redrafts under my direction, always through many iterations — restructuring, cutting, rewriting sections that don't hold up. I fact-check every claim before publication. I decide what ships and what doesn't. Pieces that don't meet that bar are discarded.

Images

Thumbnail images accompanying my articles are AI-generated, typically used as-is or with only minor adjustments. I choose the image, decide whether it fits the piece, and discard anything that doesn't.

On watermarking

Some AI companies now embed invisible watermarks in AI-generated text, partly in response to EU transparency rules. Whether a specific piece of mine carries that mark depends on how much I edited it afterward, and possibly on which model generated the draft — not on how much of the thinking, structure, or judgment was mine.

So I'm not relying on the watermark to do my disclosure, in either direction. I tell you directly that I use artificial intelligence in my process — I believe human judgment assisted by AI, used with rigor, produces better work than either alone.

This policy applies uniformly, regardless of what any given watermark check would show.

If you have questions about how a specific piece was produced, I'm glad to answer them directly.