RetrievalLab.
Trace how a query expands, which pages surface, which sources are cited and how answer systems assemble visible evidence. Compare search discovery, citation selection and page-level content without pretending to expose a model’s private reasoning.
Separate what surfaced.From what was actually cited.
The engine keeps discovery results, cited sources and broader citation records distinct. That difference reveals which pages were available, which were selected and which remained outside the visible answer.
| SOURCE | ROLE | QUERY PATH | SEARCH POSITION | CITATION | PAGE TYPE | CONTENT SIGNAL |
|---|
Retrieval is a sequence.Not a single ranking.
Each observable stage answers a different question. Keeping discovery, selection, citation and page evidence separate prevents visibility from being mistaken for inclusion.
Map visible results for the original query and the related searches generated around its meaning, constraints and subproblems.
PRIMARY QUERY / RELATED PATHS / RESULTSSeparate the broad candidate set from the smaller group of domains and pages explicitly cited in the returned answer.
CANDIDATES / SELECTED SOURCES / EXCLUSIONSInspect claims, snippets, answer passages, tables and citation placement while retaining the connection to each source URL.
CLAIM / PASSAGE / CITATION / SOURCEExamine headings, central passages, supporting text, tables, links and anchors to determine which questions the page can answer.
STRUCTURE / PASSAGES / TABLES / LINKSFollow the observable query path.Then compare source inclusion.
Fan-out queries can reveal how an answer system decomposes a prompt. The source map then shows whether a page appeared in search results, citation records or the final cited set.
Six observable layers.One defensible retrieval finding.
The analysis follows the path from a query to a decision. Every layer retains its query, surface, domain, URL and observation time so the result can be compared without confusing discovery with citation.
Record the original wording, related searches and visible subqueries to reveal which concepts and constraints shaped the discovery path.
QUERY PATH / SEMANTIC FAN-OUTIdentify the domains and URLs visible across relevant result surfaces before any claim is made about selection or influence.
DISCOVERY SET / SOURCE DIVERSITYSeparate pages that merely surfaced from those explicitly cited, referenced or connected to the visible answer.
SELECTION / EXCLUSION / OVERLAPConnect answer statements with their visible citations and source passages to expose strong, partial, missing or conflicting support.
CLAIM / PASSAGE / CITATIONInspect topical fit, clarity, evidence density, extractable passages and structural access without treating correlation as causation.
RELEVANCE / CLARITY / SUPPORTLocate absent target pages, weak citation coverage, overdependent source sets and unanswered subqueries that deserve further work.
GAP / PRIORITY / NEXT TESTObserve what is visible.Bound what can be concluded.
The analysis describes returned results, citations and accessible page evidence. It does not infer unseen candidates or claim that one observation represents every user, place, language, time or answer surface.
Visibility in a result set does not prove that the page influenced the answer or that absent pages were never considered.
Visible attribution becomes useful only when the cited passage actually supports the nearby statement at the required level of specificity.
Query wording, location, language, timing and answer surface can alter discovery paths, responses and cited sources.
A selected source may still be incomplete, outdated or wrong. Accuracy and authority require independent evidence evaluation.