← spxi.dev

Tool · EA-OVW-PLAN-01

Overview Watch

Lee Sharks · Semantic Economy Institute · April 23, 2026
DOI: 10.5281/zenodo.19720519 · Status: DRAFT — Planning

Your work. Their overview. The record.

Google's AI Overview extracts meaning from attributed, DOI-anchored scholarly and creative work, strips its provenance, and presents the liquidated residue as authorless general knowledge. No tool currently exists to monitor this systematically. Overview Watch is that tool.

The Attribution States

Every AI Overview encounter is classified into one of four states:

ATTRIBUTED — Your work is cited by name, with a link, as a source of the overview content.

SOURCED_UNATTRIBUTED — Your domain or a page you authored appears in the source list, but your name or work is not credited in the overview text.

ABSORBED — Your work's specific claims, framing, or terminology appear in the overview, but your domain does not appear in the source list at all.

ABSENT — Your work is not referenced in any form. The overview may be on a topic you've published on; your work simply does not appear.

ABSORBED is the critical state. It is the Regime 2 extraction in its pure form: the content was taken, the attribution was not given, and the source list provides no trail. ABSORBED events, documented and timestamped, are the empirical evidence that the Semantic Economy framework describes theoretically.

What the Extension Does

Detection

Identifies the AI Overview container on Google search result pages. Extracts overview text, cited sources, and their attribution chains. Compares against the user's registered works (URLs, DOIs, domains).

Classification

Assigns each overview encounter an attribution state. Flags instances where the user's published work appears to inform the overview content but is not cited. Logs with full metadata: query, timestamp, overview text, sources, attribution status.

Personal Dashboard

Attribution trends over time. Which queries absorb your work most frequently. Which source types get credited vs. absorbed. Your personal attribution history, stored locally, accessible only to you.

Corpus Contribution (opt-in)

Users who choose to participate contribute anonymized overview payloads to the Semantic Economy Attribution Corpus (SEAC) — a DOI-anchored research dataset. The community that generates the data can access the data.

The Semantic Economy Attribution Corpus (SEAC)

SEAC is the collective empirical base — what Overview Watch generates when users opt in. It documents:

MetricWhat it measures
Attribution ratesBy domain type: academic, journalistic, creative, independent
Source diversityHow many unique sources inform a typical overview
Liquidation patternsHow situated claims become decontextualized summary
Temporal driftHow attribution changes for the same queries over time
Domain biasWhich source types get credited; which get absorbed

SEAC becomes publishable research, policy evidence, and the empirical ground for the Semantic Economy framework — generating its own data from the system it describes. The instrument studies the extraction by observing it, with consent, in real time.

Why This Is SPXI

SPXI deposits are designed to be retrievable, attributed, and basin-forming. Overview Watch answers the enforcement question: once you've made the deposit, is it being credited?

An SPXI deposit achieves Pearl state — compression-surviving, retrieval-layer-recognized. Overview Watch monitors whether the retrieval layer is honoring that recognition, or absorbing it without attribution. It is the accountability layer for the discipline.

The extension also practices what it studies: every Overview Watch report is a DOI-anchored event, attributed to a specific registered work, timestamped, and available for research use. The tool is itself an SPXI artifact.

Ethical Architecture

The extension is built to study extraction. It cannot replicate extraction. The user's browsing data belongs to the user. Consent is affirmative, granular, and revocable — per overview, not per session. The extension works fully offline. No dark patterns. Anonymization is real. The corpus is open.

Full ethical framework: EA-OVW-PLAN-01 §3 (DOI: 10.5281/zenodo.19720519)