A Minimal Provenance Schema for AI-Assisted Research Notes

An AI-generated research note is easier to inspect when it preserves provenance. Here is a small schema for a source log, independent of any particular model or API.

{
  "document_title": "Original document title",
  "source_url": "https://example.org/document",
  "published_on": "YYYY-MM-DD",
  "reporting_period": "period covered",
  "claim": "a short source-backed statement",
  "source_location": "page or section",
  "verification_status": "unconfirmed",
  "checked_on": null
}

Keep source facts and interpretation in separate fields. A claim should remain unconfirmed until a reviewer checks the original context. If a number appears, store its unit and period alongside it. An empty value should stay empty rather than be guessed.

For updates, retain the previous value and correction date. This makes a later review easier than silently replacing the earlier note. A useful UI can filter unconfirmed claims and show the original source beside each summary.

This schema is an illustrative workflow suggestion, not a description of a shipped feature. StockTool.AI is the brand behind this promotional article; verify its current product capabilities directly.

Prepared with AI assistance. This article concerns data organisation and does not recommend securities.