Evidence type
Repository, CI and public-runtime evidence
Technical SEO · AEO · GEO · First-party implementation
Building D2's Search & AI Visibility Operating System
D2 converted its own website into a controlled Search & AI Visibility implementation: freeze a baseline, fix canonical and release contracts, standardize entity/author/evidence relationships, define market and Search intent architecture, publish citation-ready tools, and expose the implementation itself as first-party proof. The case demonstrates implementation and readiness evidence; it does not claim ranking, traffic, AI citation or revenue lift that has not yet been measured after production release.
Case version 2026.09 · Evidence snapshot 2026-09-03
Proof summary
Read the evidence before reading the outcome.
Evidence status
First-party implementation case · measured source-state outcomes
Measurement / operating scope
D2 website · frozen baseline, canonical/entity/evidence controls and measurable release-state implementation
Observed state
Baseline, interventions and deterministic source-state/release outcomes are published with repository/CI/runtime evidence.
Claim boundary
Implementation and readiness evidence does not prove future ranking, impressions, clicks, AI citations, leads or revenue.
Proof reviewed
2026-09-25
Case evidence contract
Baseline, action, period, measurement, outcome and limits are kept separate.
Baseline
The frozen baseline had 168 repo-owned EN/VI canonicals plus two published dynamic Insights, a failing CI quality gate, and three tracked canonical/release blockers.
Action
D2 hardened canonical and release contracts, standardized entity/author/evidence relationships, defined market and Search intent architecture, versioned citation-ready tools and exposed the implementation itself as first-party proof.
Period
Frozen baseline and measured source-state implementation reviewed on 3 September 2026, with later remediation phases recorded as separate release states.
Measurement
Repository, CI and public-runtime evidence: deterministic contract passes, production build/typecheck/lint state, Search intent architecture and versioned citation-tool inventory.
Observed outcome
The implementation produced measured source-state and release-readiness outcomes, including passing deterministic contracts and a controlled Search authority/evidence graph.
Limitations
No Google ranking lift, impression/click growth, AI citation lift, lead uplift or revenue impact is attributed until those outcomes have their own observation windows and attribution data.
Problem
A larger site is not automatically a stronger Search system.
D2 already had a broad bilingual website, service pages, Insights, tools and first-party Work evidence, but Search/GEO/AEO readiness was uneven: release gates were brittle, canonical-host behavior had defects, Search proof had no dedicated Work layer and authority assets were not yet connected through one measurable evidence model.
Phase 0 baseline
Freeze the evidence before optimizing.
84 EN + 84 VI
Static canonical inventory at the frozen Phase 0 baseline: 168 repo-owned public canonicals with full static EN/VI parity.
docs/SEO_GEO_AEO_PHASE_0_BASELINE_2026-09-03.md
2 published
Two published dynamic Insights were live at baseline, bringing the expected public indexable union to 170 URLs.
Phase 0 dynamic content baseline
CI failed
The baseline Site Quality run failed at a brittle business-architecture verifier and skipped downstream release checks.
Site Quality run 33660205175
3 tracked blockers
The frozen baseline carried a broken www canonical-host path, HTTP leakage in application redirects and the CI quality-gate failure.
P0-HOST-001 · P0-SCHEME-001 · P0-CI-001
Intervention ledger
Thirteen phases, with each phase changing a testable control.
Phase 01–03
Technical and entity foundation
Harden canonical redirect construction, normalize redirect scheme at the runtime boundary, replace stale exact-string CI assertions with derived invariants, and standardize D2's organization/service entity graph without changing stable IDs.
Phase 04–05
Author, evidence and citation worthiness
Create dedicated Commerce, Automation and Search author entities; attach provenance, evidence and claim boundaries; strengthen priority pages so visible claims and machine-readable relationships use the same evidence discipline.
Phase 06–07
Market and Search authority architecture
Keep English unprefixed as global/x-default, preserve Vietnamese under /vi, avoid cloning the site under /us, and define a controlled Search authority map with 12 intents across 5 tracks that only becomes public when content readiness passes.
Phase 08–11
Topical authority expansion
Deepen first-party authority around U.S. Commerce, GMV Max, production n8n and Vietnam Commerce while preserving evidence and market boundaries instead of scaling generic pages.
Phase 12
Tools become citation assets
Version eight deterministic tools with formulas/rules, input-output contracts, limitations, worked examples, stable #methodology anchors, copyable citations and bilingual CreativeWork methodology schema across 16 public tool URLs.
Phase 13
Expose the implementation as first-party proof
Add Search Work as a third evidence line and publish this bounded implementation report so Search service, Search Insights, Search OS, Methodology and proof all connect through one explicit authority graph.
Measured outcomes
Claim source-state outcomes first; Search outcomes require measurement.
PASS
Phase 0–12 deterministic contracts
On the final Phase 12 main head, deterministic source contracts through Phase 12 passed together before browser tooling was intentionally stopped at the agreed core gate.
main 54619ee5df308c8cd410ff548a62e262cfb443db
PASS
Production build
The production build passed on the final Phase 12 source state.
Site Quality run 33745027698
PASS
TypeScript + lint
TypeScript and lint passed on the same final main head after formatting-only cleanup of the Phase 12 files.
Site Quality run 33745027698
12 / 5
Search intent architecture
The Phase 7 planning layer defines 12 Search intents across 5 authority tracks, with publication gated by readiness rather than route generation alone.
src/lib/search-authority-topics.ts
8 / 16
Citation-ready tools
Eight deterministic tools share one versioned citation contract across 16 English/Vietnamese public URLs.
src/lib/tool-citation-assets.ts
Claim boundary
What this case does not prove.
The evidence supports architecture, source-state and release-quality claims recorded on 2026-09-03. It does not prove future ranking, impressions, clicks, AI citations, leads or revenue. Production activation of later phases still depends on rebuilding the deployed image from the corresponding main commit.
No claim is made here about post-release Google ranking lift, impressions, clicks or non-brand query growth.
No claim is made that an AI system cited D2 more often because of these changes; citation measurement requires an observation window and repeatable query set.
No lead, pipeline or revenue uplift is attributed to Search/GEO/AEO until attribution data exists for the relevant period and release state.
At the Phase 12 verification point, the public site still exposed the old immutable production image, so later source-state improvements were not treated as live production outcomes.
Authority graph
Proof should resolve back to service, knowledge, system and methodology.
Cite this case
D2 Group. “Building D2's Search & AI Visibility Operating System.” First-party implementation case, version 2026.09, reviewed 2026-09-03. https://d2group.co/work/search/d2-search-ai-visibility-operating-system