How it works
Fill in the four slots of an evidence block: claim, metric, named source and year. The builder stitches them into the exact markdown shape ChatGPT and Perplexity prefer - then scores the whole block out of 100 so you can see which entries need a named source or a year to hit full citation weight.
What a lift-ready evidence line looks like
- Bold metric first. “3-5x”, “41%”, “$499” - the first thing a scanner (human or model) reads.
- One-sentence claim. No adjectives, no fluff. The claim should be the thing a researcher would bookmark.
- Named source. Unnamed claims get skipped. Anonymised studies get skipped.
- Year tag. Recency is a trust signal. Models penalise undated claims.
Ship it
Pair with the Citation Builder to format full citations, the Fact Block Generator for standalone numeric snippets, and the Fact Density Analyzer to measure the overall page. More context in how AI assistants choose citations.