Diana Reeves has been a journalist for twenty-six years. She covered the aftermath of Hurricane Katrina from a borrowed desk in Baton Rouge. She spent three months embedded with troops in Afghanistan. She won a regional press award for a series on municipal corruption that led to the resignation of a city manager. None of that mattered when the algorithm said her writing wasn't real.
It started with a routine policy change. In January 2026, the news organization where Reeves had worked for the past eight years announced that all submitted copy would be screened through an AI detection tool before publication. The policy was framed as a quality assurance measure - a way to ensure that the publication's reputation for original reporting remained intact in an era of AI-generated content farms.
Reeves understood the logic. She had seen the content farms herself - websites that churned out hundreds of articles a day using AI, clogging search results with plausible-sounding nonsense. She supported the idea of protecting original journalism. What she didn't anticipate was that the tool would flag her own work.
The Flag
The first article to trigger the detector was a 1,200-word piece on a school board budget dispute - the kind of bread-and-butter local journalism Reeves had been writing for decades. The tool assigned it a 73% AI probability score. Her editor, a colleague she had worked with for six years, sent her a message: "Hey - the new system flagged your school board piece. Can you verify this is original?"
The question itself was the insult. Not hostile, not accusatory, but carrying an unmistakable implication: we need you to prove you wrote this. After twenty-six years, after every late night and early morning and difficult phone call that goes into building a career in journalism, the question landed like a slap.
The Pattern
Reeves verified the piece, submitted her notes, and moved on. Then it happened again. And again. Over the next two months, five of her articles were flagged. The scores ranged from 58% to 81%. Each time, she had to pause her reporting to compile evidence of authorship - interview recordings, source contact logs, draft timestamps, the accumulated proof of real work.
The pattern emerged gradually. Reeves writes in a particular style: clean, declarative sentences with minimal ornamentation. Her leads are tight. Her transitions are seamless. Her paragraphs follow a logical structure honed by decades of deadline writing. This prose - the product of a career's worth of skill - looked, to the algorithm, like something a machine would produce. She was being penalized for being good at her job.
The Conversation
Reeves eventually requested a meeting with her editor and the publication's managing editor to discuss the pattern. She presented her argument simply: the detection tool was not designed to evaluate experienced professional journalism, its training data likely underrepresented long-form local reporting, and the repeated false flags were costing her reporting time and undermining her professional standing.
The managing editor was sympathetic. But the policy remained in place. "We can't make exceptions," he told her. "If we exempt one reporter, we have to exempt everyone, and then the policy has no teeth." The logic was institutional, not personal. It was also, Reeves felt, exactly backward: the policy was supposed to protect good journalism, not interrogate it.
What Changed
After four months, the publication quietly adjusted its detection threshold from 50% to 80%, reducing the number of false flags across the newsroom. They also implemented a "known author" bypass for reporters with more than two years of tenure. Reeves's articles stopped being flagged. The policy still exists, but it bothers her less now - mostly because she knows what it says about the state of her profession.
"The question isn't whether AI detection has a role in journalism," she told me. "Maybe it does, for screening freelance submissions from unknown writers. The question is what it means when a newsroom needs a machine to tell it whether its own reporters are real. That's not a technology problem. That's a trust problem."