Newsrooms Are Putting Reporters' Bylines on AI-Rewritten Stories. Research Says That's the Wrong Fix

Quick answer

At the Sacramento Bee, more than 30 of 40 union journalists are withholding their bylines from articles rewritten by their parent company's new AI tool. The tool, called the Content Scaling Agent, takes a reporter's published story and generates new versions aimed at different audiences, then attaches that reporter's byline to the result even when they never touched the rewrite. Reporters call it borrowing their credibility to sell an AI product readers can't tell apart from their own reporting. A 2,145-person academic study published this year in Digital Journalism backs up why that specific move backfires: audiences don't punish newsrooms much for using AI on routine tasks, but they punish them hard for hiding it or skipping human oversight, and a byline that implies a reporter wrote something they didn't is exactly that kind of hidden gap.

What actually happened at the Sacramento Bee

McClatchy, which owns the Sacramento Bee, the Fresno Bee, and other California papers, built an internal tool it calls the Content Scaling Agent. CapRadio reported in April 2026 that the tool takes an already-published story and generates a second version angled at a different audience, for example reframing a school-district story for parents specifically. That new version publishes with the original reporter's byline attached, whether or not the reporter wrote a word of it.

Ariane Lange, an investigative reporter at the Bee since 2021 and vice chair of the Sacramento Bee News Guild, put the problem plainly to CapRadio: "They know the public trusts the reporters. They're banking on using our credibility as reporters to shore up the credibility of this AI tool that reporters of the Sacramento Bee do not believe in." More than 30 of the paper's 40 journalists have been withholding their bylines from AI-rewritten pieces in protest since the tool rolled out.

The dispute didn't stay theoretical. In September 2026, the New York Post reported McClatchy laid off dozens of journalists across its California papers, including at least 10 at the Sacramento Bee, the same week it added three editors to its AI-focused Content Innovation Lab. CEO Tony Hunter told staff in a leaked town hall recording that the company has "a company-wide effort to agentify the entire enterprise." Reporters kept working the traffic-fatality beat and the city-hall beat with fewer colleagues, while the company scaled up the tool putting their names on work they hadn't written.

What the research actually measured

The Bee dispute is one company's fight. A separate, unrelated academic study gives it a broader empirical backing. Sebastián Valenzuela, Ingrid Bachmann, Porismita Borah, and Natalia Solís Valdés ran a pre-registered choice-based conjoint experiment with 2,145 participants in Chile, published in Digital Journalism in March 2026 under a CC BY open-access license. They showed participants pairs of hypothetical news outlets that varied across seven dimensions of AI use, including content creation, personalization, human oversight, and disclosure, then asked which outlet they found more credible and which they'd choose to read.

Two findings stand out. First, human oversight and disclosure were the strongest positive predictors of both credibility and outlet choice, by a wide margin over every other factor tested. Second, using AI for menial tasks like transcription or spelling correction barely moved the needle either way, while AI-generated content production modestly hurt both credibility and selection, especially for harder news topics like politics or the economy. Valenzuela told LatAm Journalism Review the point wasn't whether AI got used. It was whether the newsroom stayed honest about it: "What we wondered was whether journalism, in trying to improve by using more efficient tools such as AI, might end up undermining its credibility."

Run that framework against the Bee's Content Scaling Agent and the mismatch is obvious. The tool skips disclosure by attaching a real byline to AI-generated text, and it substitutes a system-level review for the specific reporter's own judgment on a story with their name on it. Both are the exact two levers the study found readers care about most.

Readers want labels, but labels aren't simple

A companion study helps explain why McClatchy reached for a byline instead of a label. Jessica Zier and Nicholas Diakopoulos published interview-based research, also in Digital Journalism, on what readers actually want from AI disclosure labels. Summarized by Nieman Lab in June 2026, their participants wanted labeling for real reasons: keeping journalists accountable, avoiding fraud, and knowing when to double-check a claim themselves. But labels aren't free of cost. Some participants said seeing an "AI-generated" tag made them assume they needed to go find the story somewhere else to verify it. Others saw AI authorship as evidence the outlet was cutting corners, since, as one put it, "you don't need the training for that if you're going to use AI to generate your entire article."

That tension is real. A visible AI label can cost an outlet the reader's trust outright. A byline with no label costs the outlet nothing up front, but only because the reader has no way to know there's anything to distrust. McClatchy chose the option that's invisible to readers. The Bee's own reporters are the ones objecting, specifically because it isn't invisible to them.

What you can actually check

If a byline looks off (a story that reads differently from a reporter's usual voice, oddly formatted for a specific audience segment, published without much reporting texture around a topic that would normally need sourcing) there's no public disclosure requirement forcing an outlet to tell you which parts were AI-touched. The tools available to a reader are the same ones we've covered when newsrooms started running AI detectors on their own opinion submissions: paste the suspect text into a detector yourself and treat the score as a reason to ask the outlet a direct question, not as proof of anything on its own. Voluntary detector checks aren't a substitute for a newsroom's own disclosure policy. They're what's left for a reader when that policy doesn't require the outlet to tell you upfront, which right now covers most AI use in journalism. Nothing in current state or federal AI transparency law reaches ordinary published text at all, so a masthead's internal policy is the only thing standing between a byline and a bot.

FAQ

Does McClatchy's tool fabricate facts? Reporters told CapRadio the tool has produced errors during rollout, but the company's AI policy states it doesn't use hallucination-prone generation and that human editors review all output before publication. The dispute isn't about factual accuracy. It's about whose name goes on the result.

Is attaching a reporter's byline to AI-rewritten content illegal? Not currently. No state or federal AI transparency law requires disclosure on ordinary news text, so a byline policy like this is a company decision, not a legal violation, unless a specific outlet's own ethics policy prohibits it.

Does the research say readers want AI banned from newsrooms entirely? No. The Chile study found routine AI use for transcription or spelling checks barely affected credibility. What moved the numbers was hiding AI-generated content production or skipping human oversight on stories that need judgment, which is a narrower target than "no AI at all."

Run the check yourself

You can't force a masthead to disclose what a byline actually means. You can run any text you're unsure about through TheChecker.AI's free detector and use the result as the start of a conversation with the outlet, the same honest-broker approach the research above says actually rebuilds trust: disclosure and human oversight, not a name attached to work nobody signed off on.