How curaft research feeds are calculated

Each topic starts with one or more plain-language semantic queries. Results are requested from the curaft discovery backend with a thirty-day freshness window unless a product surface explicitly selects another range. Semantic matching can return relevant wording that does not repeat the query exactly.

A feed is a discovery sample. It is not a complete census of the web, and absence from the result set does not prove that a story or source does not exist.

The sample size is the number of stories loaded into the server-rendered topic page. The total matching count is taken from the backend response when available. Coverage dates are the earliest and latest valid publication dates in the loaded sample.

Dates describe the sample currently shown. They are not used to claim publishing frequency or market momentum.

Source counts group the loaded stories by their normalized source name. Recurring themes group cleaned source tags and show only themes that appear at least twice in the sample. Generic labels such as news, article, update, and trends are removed.

These are descriptive counts. curaft does not convert them into market share, sentiment, authority, or statistical significance.

A story page is eligible for search indexing only after its complete brief is available and passes automatic quality checks. The source text must be readable and substantial, the brief must contain enough original explanatory material, and it must include concrete takeaways and specific tags.

Pending, unavailable, fallback, or short briefs use noindex and are excluded from the sitemap. Indexed briefs retain the original source link and identify that source in page metadata.

The original publication remains the authority for quotes, figures, claims, and later corrections. curaft briefs are designed for discovery and triage. Readers should open the original source before citing a claim or making a material decision from it.