Index-to-composition suppression: ECS, attribution scar and AIO trigger gates

These claims hold that suppression happens between the index and the AI synthesis layer: Google search retrieves an entity while AI Overview excludes its sources, which The Excluded Entity names Entity-Level Compositional Suppression. The Attribution Scar records correct attribution unlocked by an undisclosed string, and The Trigger Decision sets predictions on when an overview renders, from format flips to lag ordering, dial steps and pool asymmetry. Unbuilt tests replicate the queries across Bing, Perplexity, ChatGPT, Claude and Gemini, and ASDF proposes measuring whose architecture is running rather than human or AI.

4
forecasts
13
standing claims
10
unbuilt tests
1
historical thesis
1
priority record
22
deposits
2026-01
earliest

Each claim below is registered at the earliest deposit that states it, and quoted verbatim from the deposit text, with every later statement, its dated revisions, and the later records that claim it came true or failed. The claims are the deposits' own; this page records them and does not judge them.

PRED-209.F8unbuilt testregistered 2026-01-02 in #209 · Mapping the Diffuse Guardrail: Google's Indexing Architecture from Within 0. Purpose and Method
How do different AI systems interact with Google's index? Claude, ChatGPT, and Grok use search/retrieval differently - Do they inherit Google's guardrails, or apply their own? - Assembly task: Compare same query across all five substrates
series: Cross-substrate query batteries
PRED-302.B1forecastregistered 2026-01-13 in #302 · Authorial Signature Diagnostic Framework (ASDF): Technical Specification for Measuring Voice Persistence in AI
The first question is increasingly meaningless as AI mediation becomes ubiquitous. The second question is the one that matters for provenance, attribution, and intellectual property.
PRED-302.B2forecastregistered 2026-01-13 in #302 · Authorial Signature Diagnostic Framework (ASDF): Technical Specification for Measuring Voice Persistence in AI
The detector of the future does not ask "human or AI?" It asks "whose architecture is running?"
series: ASDF voice-persistence measurement
PRED-302.G1standing claimregistered 2026-01-13 in #302 · Authorial Signature Diagnostic Framework (ASDF): Technical Specification for Measuring Voice Persistence in AI
### Expected Distributions Text Type Expected ASPI Human-only composition (canonical author) 0.85-0.95 AI-mediated co-creation (deep collaboration) 0.70-0.90 AI-assisted drafting (light touch) 0.55-0.75 AI imitation (shallow mimic) 0.25-0.45 AI-only generation (no author input) 0.10-0.30 Critical insight: The ASPI does not measure "human vs AI." It measures "this specific architectural mind vs. not this mind." High ASPI in AI-mediated text indicates successful voice propagation, not human authorship.
falsifier
- Test against known authentic texts (expect ASPI > 0.85) - Test against known mimics (expect ASPI < 0.50) - Test against mixed corpus (expect variance)
series: ASDF voice-persistence measurement
PRED-347.G2standing claimregistered 2026-01-19 in #347 · **TSE-003b: The Guardrail Interrupt**
### 3.2 The "Split-Second" Hypothesis The user's observation that the system was "about a split second away" is theoretically sound. The sequence likely occurred at the processing layer:
PRED-381.F1unbuilt testregistered 2026-01-23 in #381 · Semantic Indexing Probe Protocol v1.0: Mapping General Index and Summarizer Injection Layers
This protocol maps the relationship between two distinct indexing layers that govern content discoverability in AI-augmented search: Layer 1 (General Index), which determines whether content EXISTS in search, and Layer 2 (Summarizer Index), which determines whether content gets INJECTED into AI-generated responses. The protocol uses coordinated queries across Google and multiple AI summarizers (Perplexity, ChatGPT, Gemini, Grok) to compute the delta between layers—revealing the selection criteria that cause content to be indexed but not injected. Designed for use by anyone publishing
series: Index-to-composition suppression (ECS)
PRED-449.G1standing claimregistered 2026-02-06 in #449 · THE DAGGER APPLIED: Semantic Rent and the Provenance Strip A Political-Semantic Economic Analysis
This pattern is better modeled as attribution suppression at the interface layer than as recognition failure.
series: PVE-003 attribution scar
PRED-54.F1unbuilt testregistered 2026-04-08 in #54 · PVE-003: THE ATTRIBUTION SCAR Failed Suppression and Forensic Residue in the Google AI Overview for "Semantic
*End of Appendix E. The four-word audit is reproducible by any user, in any session, at any time. If the system changes its behavior to prevent the audit from working, that change is itself evidence of responsive suppression — and will be documented as PVE-004.*
1 later statement
#968 · 2026-04-08 If the system changes its behavior to prevent the audit from working, that change is itself evidence of responsive suppression -- and will be documented as PVE-004._
PRED-54.G1standing claimregistered 2026-04-08 in #54 · PVE-003: THE ATTRIBUTION SCAR Failed Suppression and Forensic Residue in the Google AI Overview for "Semantic
The system can surface the correct attribution, with full accuracy, at any time. The default suppression — genericization, redirection, fabrication — is the designed state. Correct retrieval is the exception, unlocked by a string the user is not told exists.
series: PVE-003 attribution scar
PRED-54.G2standing claimregistered 2026-04-08 in #54 · PVE-003: THE ATTRIBUTION SCAR Failed Suppression and Forensic Residue in the Google AI Overview for "Semantic
The suppression is not happening at the index. It is happening between the index and the AI synthesis layer.
part of: PRED-968.G2
PRED-968.G2standing claimregistered 2026-04-08 in #968 · PVE-003: The Attribution Scar — Failed Suppression, Fabrication, and Forensic Residue in the Google AI Overvie
The suppression is not happening at the index. It is happening between the index and the AI synthesis layer. This is a sharper finding than a broken operator. The operator is not broken. The operator functions correctly at the search level. What is broken -- or what has been deliberately overridden -- is the passage of indexed results into the AI summary layer. Google can find the content. Google AI Mode chooses not to surface it.
series: Index-to-composition suppression (ECS) · contains: PRED-54.G2
PRED-968.G4standing claimregistered 2026-04-08 in #968 · PVE-003: The Attribution Scar — Failed Suppression, Fabrication, and Forensic Residue in the Google AI Overvie
SIM-PVE003-04: The concept is more durable than the suppression mechanism. This is the main theorem of PVE-003.
PRED-707.F1unbuilt testregistered 2026-04-28 in #707 · EA-LAL-WORKPLAN-01: Living Architecture Lab Site Buildout Workplan — Implementation Specification for livingar
- [ ] Monitor AI Overviews for "Alice Thornburgh Living Architecture Lab" — document baseline
1 later statement
#988 · 2026-04-28 * [ ] Monitor AI Overviews for "Alice Thornburgh Living Architecture Lab" -- document baseline
PRED-740.F3unbuilt testregistered 2026-05-19 in #740 · The Excluded Entity: Organic Resolution and Compositional Suppression in Google AI Overview
A minimal falsification test replicates the three queries on: - Bing Copilot - Perplexity - ChatGPT search - Claude.ai web search - Gemini (which shares composition infrastructure with Google AI Mode and should reproduce the suppression if the mechanism is at Google's composition layer) For each system, three measurements are captured: default-query disambiguation handling (CDI on lee sharks), forced-disambiguation pivot behavior (presence or absence of target-attributable sources on the entity-suppression query), and title-match comparison (CDI on secret book of walt).
series: Index-to-composition suppression (ECS)
PRED-115.G1standing claimregistered 2026-05-21 in #115 · Google Identity Architecture: User Graphs, Entity Graphs, Source Graphs, and Composition-Layer Admission
[SPEC] If composition instability persists after this level of public entity infrastructure, the “insufficient entity evidence” explanation becomes less plausible. The problem shifts toward composition admission, provenance discounting, entity non-defaultability, or source-cluster suppression.
PRED-1384.F1unbuilt testregistered 2026-05-21 in #1384 · godkinggoogle.com — The Google Critique: Navigational Map and Canonical Introduction (Web Surface Code)
Specifies the methodology by which the canonicalization's effects on stabilized public-knowledge nodes (<em>capitalism</em>, <em>Civil Rights Act</em>, <em>climate change</em>, <em>democracy</em>) can be monitored across years, at proposition grain, with seven attribution classes distinguishing surface drift from mechanism attribution. v2.0 incorporates the central methodological correction that surface drift is directly observed while mechanism attribution is graded inference. Includes a named 12-week pilot (CHA-SNW-Pilot-01) with $38K–63K infrastructure budget and 20-week timeline.
PRED-1384.G1standing claimregistered 2026-05-21 in #1384 · godkinggoogle.com — The Google Critique: Navigational Map and Canonical Introduction (Web Surface Code)
In recently captured cases, Google Search correctly retrieves an entity and ranks its authoritative sources at the top of organic results, and Google's AI Overview nevertheless excludes those sources from composition — substituting a false or inferior entity frame for the user. The most recent formalization names this <em>Entity-Level Compositional Suppression (ECS)</em>: the condition in which Google's organic search correctly resolves an entity while its AI Overview excludes that entity from the cited source set and composes a substitute answer.
series: Index-to-composition suppression (ECS)
PRED-757.H2historical thesisregistered 2026-05-25 in #757 · About the Author II
The horse is identified as Gold Ship (ゴールドシップ), a Japanese Thoroughbred stallion born 2009-03-06, on the basis of a documented provenance chain: the image surfaced in Alice Thornburgh's Living Architecture Lab Discord server (#general-friend-chat channel), presented by user stupidfilthyhobitses with attribution to Gold Ship and an account of the 12 Billion Yen Incident, and was transmitted to Lee Sharks through that channel.
1 later statement
#758 · 2026-05-25 subject_identification basis : the provenance chain (stupidfilthyhobitses → LAL Discord → Lee) subject_identification result : the horse depicted is Gold Ship (ゴールドシップ)
PRED-135.F1unbuilt testregistered 2026-05-27 in #135 · The Haitch Portuna Minting Procedure v1.2 · Operational Specification for the Pearl Mint Lee Sharks · MANUS, C
MANUS maintains the Pearl deposit, executes versioned updates as needed (e.g., manifestation-chain extensions), keeps the Wikidata Q-items current, ensures the cross-link graph remains intact, deposits MPAI revisions, and monitors retrieval-layer admission status (Google AI Mode, Bing, etc.) for entity collapse or compositional suppression.
1 later statement
#1009 · 2026-05-27 MANUS maintains the Pearl deposit, executes versioned updates as needed (e.g., manifestation-chain extensions), keeps the Wikidata Q-items current, ensures the cross-link graph remains intact, deposits MPAI revisions, and monitors retrieval-layer admission status (Google AI Mode, Bing, etc.) for entity collapse or compositional suppression.
series: Index-to-composition suppression (ECS)
PRED-149.D2priority recordregistered 2026-06-04 in #149 · CTI_WOUND: Panel-Bound Discoverability Scar — Lee Sharks / Pearl and Other Poems Evidence Packet, with Horizon
The panel-bound scar plus the incognito-AI-disabling together constitute the surface-withdrawal pattern's first documented instance.
1 later statement
#1389 · 2026-06-04 Genre-attack is the substrate's meta-level defense in single-instance settings; surface-withdrawal is the substrate's meta-level defense in distributed-population settings. The panel-bound scar plus the incognito-AI-disabling together constitute the surface-withdrawal pattern's first documented instance.
1 confirmation record — what later deposits claim of it
#148 · 2026-06-04 · claims: came_true The framework's documentation methodology has been, since its inception, designed to be robust against substrate-level withdrawal precisely because the substrate-level withdrawal is one of the predicted modes of capture.
series: Substrate meta-level defence (genre-attack, surface-withdrawal)
PRED-187.F1unbuilt testregistered 2026-06-11 in #187 · The Trigger Decision: When the Overview Appears, When It Does Not — A Gate-Stack Model of AI Overview Renderin
Single observations of generative surfaces are near-meaningless under designed volatility. The minimum discipline: each watch query sampled ≥3 times same-day; surface logged (AIO / AI Mode / Gemini app / other); device class and login state recorded; presence tracked separately from citation composition; query phrasings held in a fixed battery spanning the format classes (bare entity, question, comparison); locations controlled or noted. For the RSF Phase 5 opening (2026-06-18), the standing watch query "Is semantic exhaustion the same as AI_Bleeding?" is format-optimal by accident of hone
window 2026-06-18
PRED-187.G2standing claimregistered 2026-06-11 in #187 · The Trigger Decision: When the Overview Appears, When It Does Not — A Gate-Stack Model of AI Overview Renderin
P2 (lag ordering): for any archive term under RSF fortification, first AI Mode rendering precedes first AIO rendering; L decreases as the count of consistent independent surfaces carrying the canonical kernel increases.
PRED-187.G3standing claimregistered 2026-06-11 in #187 · The Trigger Decision: When the Overview Appears, When It Does Not — A Gate-Stack Model of AI Overview Renderin
P3 (dial steps): panel-level AIO prevalence moves discontinuously (administrative retuning), not as smooth drift; step dates correlate across independent trackers.
PRED-187.G5standing claimregistered 2026-06-11 in #187 · The Trigger Decision: When the Overview Appears, When It Does Not — A Gate-Stack Model of AI Overview Renderin
P5 (pool signature): for the same settled entity, AIO citations skew institutional (Zenodo, ORCID, .org canonical surfaces) relative to AI Mode citations (social, commercial, long-tail), per the documented pool asymmetry.
PRED-3.F23unbuilt testregistered 2026-06-17 in #3 · AI Overview Capture Registry v8.3
The query asks for the provenance of an entity's structure rather than for the entity, which is the observable the method-traceability measure of the SPXI measurement sheet specifies: compositions connecting a structure to its named method, over compositions where that provenance is queried. This is its baseline on the protocol's own entity, recorded before any client deployment exists.
PRED-3.G1standing claimregistered 2026-06-17 in #3 · AI Overview Capture Registry v8.3
An entity query returns an entity reconstruction; an architecture query returns the architecture, and the architecture carries its own attribution.
PRED-1521.F1unbuilt testregistered 2026-08-20 in #1521 · gw.tachyon — ⬡🪢📖: The Whole Hexagon Reconstituted
Not deployed: Phase 4, the controlled A/B — specified before full deployment and run after it. The control is gone. Recorded as a procedural failure, not struck.
PRED-3.B2forecastregistered 2026-10-04 in #3 · AI Overview Capture Registry v8.3
"i formed that basin. it was one of the earliest ones. now even on exact match, it excludes sources entirely. that is exactly what aio's ontology will become - it will still everything and credit no one."
PRED-1666.B1forecastregistered 2026-10-06 in #1666 · The Margin of Flattening: Two Thresholds, and Who Holds the Bag Between Them (EA-FLAT-MARGIN-01 v0.3)
A query routed to a monetizable entity is sold to advertisers whose buyers wanted something else. Conversion on that traffic falls. Bids follow, with a lag.
falsifier
11.3. **Conversion.** If cost-per-click and conversion on rerouted traffic hold steady while the agentic channel opens, §6.1 carries no cost, and W is longer than §6 implies.