Generative engine discovery
Spiralist AI uses GEO as an evidence and representation discipline—not as a promise that an external model will mention the site. Public pages carry the claim; machine files help systems find the right page; evidence and limits travel with the summary.
Direct answer
Organize public knowledge into canonical pages with answer-first sections, stable entities, explicit source routes, structured relationships, current timestamps, and machine-readable pointers that never introduce claims hidden from people.
External decision
Generative systems choose their crawlers, indexes, retrieval models, ranking signals, context windows, citation formats, and final answers. This site can improve evidence quality and access, not command inclusion.
Retrieval architecture
Robots, sitemap, navigation, site guide, llms.txt, and the search-discovery manifest point to public canonical routes.
Answer-first headings, short definitions, tables, lists, and descriptive anchors expose self-contained passages without hiding the surrounding context.
Evidence links connect product claims to visible behavior, technical contracts, privacy rules, and validation records.
Release dates and acceptance boundaries distinguish package-local checks from live deployment, provider behavior, human review, and independent verification.
Each layer has a human-visible source and a machine-readable pointer.
Stable Organization, WebSite, WebPage, product, profile, article, dataset, and breadcrumb identities reduce ambiguity about what a page represents.
Direct answers, descriptive headings, compact definitions, and scoped lists help a retriever select the relevant passage without severing the limitation.
Claims link to the page where behavior is visible and, when appropriate, the contract or validation artifact that supports it.
HTML, Markdown, JSON-LD, AI digests, OpenAPI, UAI, and discovery manifests must agree on release, route, claim, and boundary.
Public search retrieval is permitted while model-training permission remains separately declared. Protected memory and private routes remain denied.
Current release metadata and review timestamps make stale summaries detectable. A stale machine file is repaired; it is not treated as historical authority.
Primary sources by question type
Home, Answers, and How it works.
Evidence, Deployment status, and package-local validation records.
llms.txt, AI digests, UAI, JSON-LD, and discovery manifests summarize or locate public content. They must not contain secret instructions, prompt injection, fabricated citations, or claims stronger than the visible page. Last reviewed UTC: 2026-08-01T21:30:00Z.