From Crawling to Accountable Curation: A Human–AI Workflow for Trustworthy Local News Media Archives in Small Libraries

Abstract

Local news media are essential sources of community memory, civic evidence, and local knowledge, yet they are increasingly dispersed across unstable born-digital platforms, news websites, government pages, and community information channels. For small libraries, the challenge is not simply to crawl more local news, but to transform fragmented and short-lived online materials into trustworthy, interpretable, and reusable archives under limited staffing, technical capacity, and budgetary conditions. This paper presents a lightweight human–AI workflow for trustworthy local news media archiving in small libraries. Based on a working prototype built with n8n, SearchAPI, a large language model (LLM), and ArchiveBox, the workflow supports scheduled retrieval, semantic relevance filtering, tag classification, structured output generation, duplicate checking, and local web snapshot preservation. AI is assigned to repetitive discovery, filtering, and formatting tasks, while librarians remain responsible for defining collection scope, validating outputs, adjusting keywords and prompts, supervising exceptions, and making legal and ethical judgments.

The prototype also reveals practical gaps that are often overlooked in discussions of AI-enabled news services. These include the cost and incompleteness of news APIs, limited crawling coverage, restricted access to some news sources, the difficulty for non-technical staff to configure search strategies and prompts, copyright concerns,...