Publishers pick pragmatic AI

Publishers spent 2023 talking about whether artificial intelligence would wreck journalism, and by April 2026 many of them are using it for much duller jobs: transcription, translation, summaries, search and internal chat tools that save reporters time without replacing reporting. Digiday’s April 8 report says the shift is from experiments to daily workflow, based on a fourth-quarter 2025 survey of 40 publisher professionals and interviews with executives running artificial intelligence projects. (digiday.com) One concrete sign of that shift is where the tools sit. Publishers are putting artificial intelligence behind the scenes, inside newsroom systems and audience products, instead of asking it to write entire newsrooms’ output from scratch. (digiday.com) (wan-ifra.org) The pattern shows up outside the United States too. World Association of News Publishers case studies published in May 2025 found Diez.md in Moldova cut article-summary work from about one hour to 10 minutes, while Baku Press Club in Azerbaijan used a generative artificial intelligence social-post tool and reported a 7% page-view increase over five months. (wan-ifra.org) That is why the mood inside publishing has become less “ban it” and more “box it in.” The same outlets testing artificial intelligence for workflow gains are also demanding human oversight, internal rules and clearer contracts about what outside models can take from their reporting. (wan-ifra.org) (digiday.com) The money fight is happening in parallel. Press Gazette reported on March 24 that Financial Times owner Nikkei, News Corp, Reach and other publishers have struck deals with artificial intelligence companies, while The New York Times, Chicago Tribune owner Alden papers, Penske Media and others are also suing or backing lawsuits over training data, search traffic and copyright. (pressgazette.co.uk) Google has made that split sharper. Press Gazette says Google’s artificial intelligence agreements announced in December 2025 were described as extensions of existing commercial partnerships rather than straightforward licensing deals, which matters because publishers want explicit payment and usage terms, not vague cooperation language. (pressgazette.co.uk) That is also the backdrop to Hachette chief executive David Shelley’s April 8 comments in Fortune Europe, where he framed the fight as one over copyright, authors’ rights and the future value of human-created work while taking on big technology platforms including Google. Fortune’s published summary describes Shelley as defending copyright and arguing over “the battle for ideas,” which is a sharper version of what many publishers are now saying in private boardrooms. (readly.com) So the industry’s working position in 2026 looks surprisingly practical. Use artificial intelligence to shave minutes or hours off repetitive tasks, keep editors in charge of anything public-facing, and make the technology companies either sign licenses or fight in court. (digiday.com) (wan-ifra.org) (pressgazette.co.uk) There is even a policy version of that same idea now moving beyond private deals. A March 9 Poynter essay pointed to proposals in Europe, Indonesia, Latin America and the World Intellectual Property Organization for statutory licensing systems that would require artificial intelligence companies to pay publishers for past and future use of news content. (poynter.org) That leaves publishers trying to do two opposite things at once, and doing both on purpose. They want artificial intelligence inside the newsroom as a faster copy desk, search assistant and production helper, and they want strict borders around the same technology when it reaches outside the building and starts ingesting journalism at scale. (digiday.com) (poynter.org)