Skip to content
Article The Lenfest Institute Sep 2026

Lenfest Institute: How the Philadelphia Inquirer's AI tool surfaces hyperlocal news leads

The Philadelphia Inquirer built Scrape, an AI tool that monitors fragmented local information sources — municipal meeting calendars, school events, social media activity — and surfaces potential story leads for hyperlocal newsletter editors. The project was supported by the Lenfest AI Collaborative and Fellowship Program.

The problem Scrape was designed to solve is one of volume and fragmentation. A single editor responsible for a neighborhood newsletter was spending roughly 12 hours a week manually scanning dozens of small, dispersed sources for anything worth covering. The monitoring work left insufficient time for the actual editorial work of evaluating and developing stories.

Scrape handles the monitoring layer. Editors now receive a curated set of leads and spend their time on story judgment rather than source watching. The tool now supports six newsletters, with expansion to eight underway, and nearly 20 journalists draw on its output as part of their regular workflow. From a pilot project, it has become infrastructure the publication depends on.

The most instructive part of the case study is how the team made the tool useful. Early versions produced vague or off-target results because the prompts treated newsworthiness as a general concept. The improvement came when the team encoded specific editorial context directly into the system: target audience demographics, geographic boundaries, relevant topic categories, and explicit coverage exclusions. The more precisely the editorial criteria were defined, the more actionable the output became.

This is a transferable lesson for any publication building AI-assisted tools for beats or newsletters with high source volumes. AI is not good at inferring what a specific publication cares about from vague instructions. It is much better when given concrete parameters — the same parameters an editor would give to a new reporter joining the beat. The design task is not prompting an AI to be smart about newsworthiness; it is encoding the newsroom’s existing editorial judgment into the tool’s instructions.

The Inquirer case is most relevant to regional and local publications, but the pattern applies wherever a beat requires monitoring many small sources for a narrowly defined audience.