Ask — probe a single prompt's retrieval

See exactly which sources the retrieval layer returns for one buyer-style question, before committing it to a corpus.

Run — score a corpus and save the result

Edit the corpus (brands you track + intent-tagged queries), label the run, and execute. A 6-query corpus takes ~20s in advanced mode; results are saved server-side and appear under Runs.

Corpus JSON

Execute

Labels: letters, digits, dot, dash, underscore. Re-run the same corpus with a new label after changing content or phrasing, then diff in Compare.

Runs — everything saved on this server

Click a run to see its full report.

Compare — the experiment loop

Pick a before and after run of the same corpus. The diff shows share-of-voice movement and every query where a brand's visibility flipped.

→