What you get back
AGeoPlan with three grounded streams:
participation_targets— the real threads/communities to engage. EachParticipationTargetcarries acommunity, a realpermalink(a verbatimsource_urlfrom a verified quote — never invented), thewhy(what that audience is asking, from a cluster claim), and a value-firstsuggested_angle.citability_probes— the conversational queries to test whether you’re cited. EachCitabilityProbe’spromptis a real question phrased the way someone would ask an answer engine, derived from a cluster claim (target_phrase). Run them against ChatGPT / Perplexity / Google AI and check for your citation.answer_briefs— the answer-first content to publish. EachAnswerBriefleads with thequestion, then a groundedanswer, and carriesstat_anchors(the cluster’s real distinct-author / mention counts) plusevidence_refsthat resolve against the report’s evidence.
The honesty contract
Participation targets and citability probes are deterministic — pulled straight from the report’s permalinks and cluster claims, never the model’s imagination. The answer briefs’ prose is LLM-authored, but the answer is a factual claim, so cite-or-die is correct here: every brief’sevidence_refs resolve against report.evidence, and a brief whose evidence doesn’t resolve is
dropped before it ships. The stat_anchors are the report’s real counts.
GEO is a compounding stream — first citations take roughly three months. Treat this as a
patient play: show up value-first in the real threads, disclose your affiliation, never drop a bare
link. metalworks plans and drafts; a human runs it — nothing here posts.