Eight features, and the reasoning behind each one

One blended score is a comfortable lie. It cannot tell you the difference between a rigorous study of an adjacent question and a persuasive post about exactly your question, and those are two completely different problems to have.

So relevance and trust stay separate all the way through the interface, in the list, in exports, and in anything you share. Sort by either. Sort by the weaker of the two when you want to know whether anything here is strong on both counts. Sources that split sharply between the two axes are marked, because the split is usually the most useful thing on the row.

A page can carry your keyword forty times and still be a shopping guide. Relevance is scored on whether the source engages the question behind the keyword, how specific it is to that question rather than the surrounding field, whether it studies the population and setting you asked about, and whether your topic is the subject of the piece or a passing mention in a paragraph about something else.

Broad overviews are marked as broad instead of being pushed down, because sometimes an overview is exactly what you want. What they stop doing is crowding out the narrow work that answers you precisely.

Suggestions arrive as queries you could have typed yourself: no proprietary operators, nothing hidden. Beside each one sits the move it represents, splitting a conflated condition, naming a population, swapping in the field's own vocabulary, putting a floor under the date, cutting a commercial cluster, and one sentence on why this search rather than yours.

Running a refinement opens it alongside the original rather than replacing it, so you keep both result sets and can see precisely what changed. Every refinement you take is kept with the search, which means weeks later you can reconstruct not only what you found but why you went looking there.

Send a result set out as CSV for a spreadsheet, as BibTeX or RIS straight into a reference manager, or as a Markdown brief that reads like a short memo: the question, the ranked sources with both scores, the gaps, and the refinements you ran.

Or share a read-only workspace link. The recipient sees the same list, the same signal breakdowns and the same gap map, including any overrides you made and the reasons you gave for them. Links can be revoked, and Team workspaces can require sign-in before a link resolves.