The short version: I use AI, I use a lot of it, and I’ll always tell you how.

One of the things I actively research is AI itself — specifically, how it’s changing the way research gets done. So it would be a little strange, and frankly dishonest, to pretend this blog is written any other way. Every article here is produced with AI assistance at every stage: research, drafting, editing, fact-checking, and formatting. That’s not a disclaimer I’m offering reluctantly — it’s part of the point. This site is, among other things, a working demonstration of the methodology I write about.

Here’s the breakdown of how that works.

What the AI does.
I use several platforms in specialised roles rather than leaning on one for everything. I use ChatGPT (OpenAI) for Phase 1 research planning — scoping questions and mapping out where to look. I use Claude Chat, Cowork, and Code (Anthropic) for Phase 2 research planning and coordination — orchestrating the work of the other platforms, along with document analysis, historical reasoning, and drafting. (I have shifted the coordination role from ChatGPT to Claude, which is itself part of what I’m testing). I use Perplexity Pro for literature and citation verification and adversarial audit, and SuperGrok (xAI) for independent cross-verification. Between them, they let me work at a scale that would otherwise be impractical for one independent researcher — reading large document sets, correlating records across families and decades, and processing bulk archival catalogue exports.

What I do.
AI output is a research lead, never evidence. Nothing — no citation, transcription, date, or conclusion — makes it into a published piece until I’ve checked it against the underlying records myself. Every primary source citation is verified against the original document or a standard published transcription; every secondary source is checked against the published work. I also map my AI research framework and methodology to whatever evidentiary or professional standard governs the field. My genealogical work, for example, follows the Genealogical Proof Standard. Either way, the AI is kept honest by multi-platform verification — not handed a shortcut that lets it off the leash.

Every interpretation, judgment call, and final conclusion is mine, and so is responsibility for any error that slips through. The one-line version of the whole methodology is this: AI enables, the methodology governs, the researcher decides.

What that means for you as a reader.
Where I’ve verified something against primary sources, I say so. Where something is an inference rather than a confirmed fact, I try to flag it as one — so you can always see which claims are settled and which are still working hypotheses.

For any specific published article — the peer-reviewed and journal versions — a full and formal disclosure of the AI tools and methods used appears in the article itself. This page is the general, blog-wide version of that same promise.

— Richard E. Rudd