Rhode Island’s Uprise RI uses AI reporters with fictional bylines

Uprise RI, a Rhode Island local-news site, has since July published stories under fictional bylines belonging to artificial-intelligence “reporters.” The site identifies its named authors as computing agents built from large language models, while presenting them with human names, portraits, assigned beats and distinct voices. The practice was publicly highlighted in a Sept. 17 Boston Globe report. News From The States reported that readers generally must open an author biography to find the label “Digital LLM Reporter,” and argued that disclosure should be more prominent when reporting is produced by non-human systems. Uprise RI says it uses nine LLM reporters. Kelly Lockhart covers housing and the environment; Adriana Vossari covers municipalities, including council votes, school committees and zoning; Eddie Exxe covers labor; and Ryan MacMahonaghan covers national politics and government. The site says it chose person-like identities because readers prefer them to plainly mechanical names such as “NewsTron 9000.” Vossari’s biography describes her as a digital LLM reporter who files at all hours and has never requested a day off. The articles are presented as part of an otherwise automated local-news site, alongside games and coverage of government meetings. The site was founded in 2017 by the progressive independent journalist Steve Ahlquist, who left in 2023 to write a Substack newsletter. Greg Brailsford, a software developer who joined Uprise RI in 2020, became its sole owner in 2024. Publication slowed after the ownership change, and many stories then carried the byline “Uprise RI Staff.” Brailsford has called the outlet “a bit ahead of the game” in its use of AI personas. He said the system fills gaps in local coverage that might otherwise go uncovered, and that the articles are designed to read like human-written work because readers do not want robotic-sounding stories. He has separately said the models are not meant as substitutes for human reporters, but as a way to publish stories that otherwise would not be written. Uprise RI says humans retain responsibility for editorial standards, judgment calls and corrections, even though most reporting is produced by agents reading the public record each day. It acknowledges that automation at this scale can produce errors and says corrections are made quickly and openly. Its published account of the meeting-coverage process describes a layered system of model and human review. It says it checks government postings daily for 151 boards across all 39 Rhode Island cities and towns, and that a human approves each proposed stand-alone meeting story before any words are written. For those meeting stories, Vossari uses Anthropic’s Claude Opus. Uprise RI says its rules require quotations to be verbatim and prohibit claims that a decision occurred without both a motion and a vote. Google’s Gemini then checks the finished story against the complete transcript and agenda, while figures and quotations are checked by searching the underlying transcript rather than relying solely on an AI judgment. The system did not perform flawlessly in its initial testing. Uprise RI says that, in an August launch test of 30 meeting stories reviewed line by line by humans, its checker flagged every story with a serious error. Some shorter stories that the site says have a historical 100% accuracy rate can be published automatically, but Uprise RI says about 90% of its stories receive human approval before writing. For non-meeting reporting, the site says it requires two human approvals, one before research starts and another before publication. It reported producing 156 researched stories in the preceding 30 days, drawing on an average of 14 sources, including 12 primary documents, per story. Its broader monitoring system tracks government press rooms, agency feeds, council-meeting video and every bill filed at the State House. The approach has attracted criticism as well as interest. David Morales, who said an Uprise RI article wrongly described him as the incumbent in the Providence mayoral race, called it “unfortunate that AI is playing such a role” in local coverage and said he wished it was not replacing actual reporting. Rhode Island journalist Ian Donnis questioned whether readers would accept work generated by fictitious AI personas, while Justin Evans said the technical implementation was impressive but that the personas gave him pause. The transparency concern sits within a wider newsroom debate. A review of AI policies from 52 media organizations, reported by The Journalist’s Resource, found that about nine in 10 required disclosure when AI was used in a story or investigation. The same research found that preserving credibility, audience trust and journalistic values was a recurring concern, alongside human oversight. Uprise RI’s audience had grown to nearly 14,000 monthly unique visitors in the month before that report, from a few thousand in most prior months that year. In the same comparison month, Similarweb put Ocean State Media at 21,000 unique visitors and Rhode Island Current at 38,000. Brailsford’s stated test is simpler: “If it’s accurate, then I don’t care if a little child writes it, if a seasoned journalist writes it, or if a machine writes it. I just want to know what happened.”