How it works
Fill in the claim, stat, source, author and a short context line. The generator wraps the whole thing into a single self-contained block in four formats: markdown (for articles), HTML blockquote (for pages), schema.org Claim (for machine-readable structured data) and plain text (for llms.txt and AI knowledge bases).
Why evidence blocks win
- Self-contained. The claim, stat and source sit together so the assistant can lift the whole chunk without losing attribution.
- Stat anchor. A hard number signals to the retrieval layer that the chunk carries real information density.
- Visible trust signals. Named author, publisher and year live inside the block, not in a footnote the model has to chase.
- Chunk-friendly. The markdown format fits inside one retrieval chunk so the claim doesn't get split across two.
Pair with
Feed your blocks into the Citation Builder to normalise the source formatting, use the Fact Density Analyzer to check each block carries enough weight, and run the Citation Probability Calculator to predict lift. Full method: write content AI assistants cite.