Triage same-name genealogy records and preserve the uncertainty
At a glance
A family historian uses Claude and ChatGPT to rank candidate records and turn research into durable notes. This workflow is listed under Genealogy, Research, Knowledge Management.
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Gather the actual records and search results yourself, keeping one chat per ancestor or research problem. Prepare candidate summaries or cropped record screenshots. 2) Paste the candidates into Claude or ChatGPT and ask it to group them as strong, weak, or unlikely matches, explaining the differences. 3) Compare dates, relationships, and chronology against the underlying records before accepting a match. Keep ambiguous readings unresolved. 4) Have the assistant write findings, open questions, and repeatable research instructions to Markdown or JSON. Review those files so the next session can resume from the evidence and remaining questions.
How it went
The author says first-pass review became several times faster and sometimes surfaced promising leads he might otherwise have skipped.
Worth knowing
He also documents a confidently wrong parish-register reading from the wrong page. Check the surrounding page and record identity, not just the cropped text.
