How a claim gets checked before it reaches your article
A reel can say anything. Here is what StrandScribe does between hearing a claim and letting it into a draft.
Short videos are a good place to find a topic and a poor place to find a fact. A confident voice and a number on screen are not evidence. So when you share a reel with StrandScribe, the reel is treated as a lead, never as a source.
First, the claims are pulled out
The audio is transcribed and reduced to a short list of checkable claims: the statements that are either true or false and that an article would stand on. Opinions, jokes, and calls to follow the account are left behind.
Then each claim is searched
When you ask for sources, each claim is searched on the open web and in public reference and research indexes. The search only produces addresses. Nothing a search engine says about a page is trusted.
Every page is read, and the passage has to be there
Each page is fetched and read. A source is kept only when a passage that bears on the claim can be found word for word on the page that was actually fetched. If the passage cannot be located in the page text, the source is dropped. That one rule removes most of the ways a research tool can mislead you: summaries that drift, quotations that were never written, pages that have changed.
A page can support a claim or dispute it. Both are kept, and both are shown to you with the exact passage, so you can see what the article will be built on before a word of it is written.
The article is written from the sources
The draft states as fact only what a kept passage says, and marks each such sentence with the number of its source. A point from the reel that no source supports is simply not in the article. The source list at the end is assembled from the kept passages themselves, not written by the model.
What you get is a piece that reads as your own reporting, with a paper trail behind every factual sentence.