Poynter: An AI company set out to fix news deserts. Instead, it copied local journalists' work
Angela Fu’s April 2026 investigation for Poynter examines what went wrong at Nota, an AI company that created a network of 11 local news sites covering underserved counties in the United States. The pitch was credible: use AI to produce civic information — government meetings, public records, community announcements — in places where local journalism no longer exists. The execution was something different.
Poynter found more than 70 stories on the Nota News sites that contained uncredited material from at least 53 journalists at 29 outlets. More than 30 copied images appeared without attribution. Some of the copied work came from outlets that were paying clients of Nota’s parent company.
The editor at the center of the findings, Jorge Rodríguez, admitted that he had used other reporters’ work and apologized. He described an operation with no clear editorial guidelines and a production quota of 10 to 15 stories per day. Nota’s CEO, Josh Brandau, said the problems stemmed from contractors failing to follow editorial standards and that each story cost under $10 to produce. The sites were quietly taken down after the investigation.
The structural failure here is specific and instructive. The issue was not that AI wrote the stories — it is that the editorial layer between source material and publication was thin to nonexistent. The incentive structure rewarded speed and volume. The oversight that would have caught copying was not in place before publication.
Several errors in the published stories were mechanical rather than plagiaristic: outdated legislation described as newly passed, paraphrases presented as direct quotes, statements misattributed. These are errors that a competent editor would catch, and they suggest that human review was either cursory or absent.
The case also illustrates a risk that comes specifically from AI-assisted production at scale: the same failure mode that produces one bad story can produce a hundred before anyone notices. Rodríguez reportedly had no process for verifying sources independently. When that gap exists, AI tools accelerate the gap, not the journalism.
For editorial teams thinking about where to place human review in an AI-assisted workflow, this investigation is a clear data point. Volume and speed gains in generation are only useful if the verification and attribution steps happen before publication, not after a scandal reveals they were missing.