I have a tendency to jot down about AI from the angle of the bleeding edge, taking a look at how journalists and media corporations are utilizing the expertise to change the way they work, reach new audiences, and transform their organizations. However the actuality is that there’s a stigma round utilizing synthetic intelligence within the journalism group. In conversations I’ve with working reporters and editors, there’s clearly nonetheless loads of reluctance, if not outright disdain, for utilizing AI in virtually any a part of their work.
Taking a look at current protection of journalists utilizing AI, nevertheless, you may suppose a few of that disdain goes away. The Wall Road Journal recently profiled how Fortune enterprise editor Nick Lichtenberg makes use of AI to turbocharge his output, typically writing as many as seven tales in a single day. The identical day, Wired highlighted how a number of outstanding reporters—together with independents like Alex Heath and Taylor Lorenz in addition to The New York Occasions’ Kevin Roose—use AI in varied editorial duties, typically within the writing itself.
With all this, it feels as if a sort of dam has burst, and I don’t suppose it’s a coincidence that it’s taking place on the identical time Claude Cowork—which brings extremely highly effective agentic AI to everybody—has remodeled the AI panorama. (An fascinating apart buried in all this protection of journalists’ use of AI is that it seems Claude is quickly changing into what the Mac grew to become amongst media professionals: the platform of selection for creatives who “know higher.”)
{“blockType”:”mv-promo-block”,”information”:{“imageDesktopUrl”:”https://photos.fastcompany.com/picture/add/f_webp,q_auto,c_fit/wp-cms-2/2025/03/media-copilot.png”,”imageMobileUrl”:”https://photos.fastcompany.com/picture/add/f_webp,q_auto,c_fit/wp-cms-2/2025/03/fe289316-bc4f-44ef-96bf-148b3d8578c1_1440x1440.png”,”eyebrow”:””,”headline”:”u003Cstrongu003ESubscribe to The Media Copilotu003C/strongu003E”,”dek”:”Need extra about how AI is altering media? By no means miss an replace from Pete Pachal by signing up for The Media Copilot. To study extra go to u003Ca href=u0022https://mediacopilot.substack.com/u0022u003Emediacopilot.substack.comu003C/au003E”,”subhed”:””,”description”:””,”ctaText”:”SIGN UP”,”ctaUrl”:”https://mediacopilot.substack.com/”,”theme”:{“bg”:”#f5f5f5″,”textual content”:”#000000″,”eyebrow”:”#9aa2aa”,”subhed”:”#ffffff”,”buttonBg”:”#000000″,”buttonHoverBg”:”#3b3f46″,”buttonText”:”#ffffff”},”imageDesktopId”:91453847,”imageMobileId”:91453848,”shareable”:false,”slug”:””,”wpCssClasses”:””}}
A cautionary story in copy and paste
Nonetheless, if the connection between journalists and AI has been warming up, it obtained splashed within the face with a chilly bucket of water final week when The New York Occasions severed ties with a contract author who had submitted a guide assessment that was at the very least partially AI-written. The review by Alex Preston, printed in early January, included passages that had been practically similar to Christobel Kent’s review of the identical guide that was printed in The Guardian months earlier.
Preston admitted he used AI to help in writing his guide assessment, saying that he had “made a critical mistake.” Whereas the incident is actually a wake-up name for the Occasions (and not necessarily the first one) about the way it communicates its AI policy to freelancers, it’s additionally a flapping crimson flag for any newsroom tempted to permit extra AI use of their operations. All of the sudden, there’s an error that appears to justify all the foundations in opposition to it.
That’s why it’s necessary to confront this immediately. The incident steers us again into the darkish cave of AI scandals in media—from CNET’s bot-authored service journalism to the made-up guide titles within the Chicago Sun-Times’ “summer reading list” last year. It threatens to undermine all of the good points many journalists and newsrooms are reaching in productivity, content material optimization, and extra, and probably encourages these simply taking their first steps with AI to fall again on the simple, blanket rule of “simply don’t use it.”
That’s why it’s necessary to look carefully at how AI was used so we will higher delineate between good and dangerous AI use. It’s simple to say there wasn’t sufficient “human within the loop” (an more and more unhelpful time period)—however the place within the loop? With prompting, fact-checking, one thing else? The entire level of AI is to outsource some human decision-making to stylish machines, so slightly than mentioning the apparent—that people must form and monitor the method—it’s higher to zero in on the particular choices that AI was requested to make, and whether or not the human gave the proper parameters and restrictions.
While you study this carefully, it undoubtedly seems the reply is not any. In accordance with The Guardian story, the 2 opinions have eerily related language—so shut that it’s troublesome to argue in opposition to outright plagiarism. Have a look at these two passages:
- Unique assessment, printed August 21, 2025: “most importantly a tune of affection to a rustic of contradictions, battered, war-torn, divided, misguided and miraculous: an Italy the place life is costume and the efficiency of artwork, and the place circuses spring up on wasteland.”
- Occasions assessment, printed January 6, 2026: “populate what’s in the end a love tune to a rustic of contradictions: battered, divided, misguided and miraculous. That is an Italy the place life is efficiency, the place circuses rise on wasteland.”
Wanting on the dates and the unquestionable similarities, we will draw some conclusions. It’s apparent Preston immediately or not directly requested the AI to create textual content he meant to incorporate within the piece, and never simply based mostly on his notes. On condition that the 2 opinions had been printed 4 months aside (and contemplating the sometimes prolonged enhancing course of on the Occasions, he probably submitted it a lot earlier), that’s virtually actually not sufficient time for the AI’s coaching information to be up to date. Which suggests the AI device he used was incorporating net search (aka RAG) to provide you with the copy.
This was a mistake. Giving Preston the good thing about the doubt, he could not have intentionally instructed the AI he was utilizing to synthesize different opinions of the guide, and maybe it grabbed The Guardian assessment by itself. However he actually didn’t inform the AI not to do this, which might appear to be a necessary a part of your immediate if you wish to keep away from the very plagiarized textual content he ended up together with.
From taboo to device
It bears repeating: In lots of—if not most—instances, how you utilize AI issues greater than whether or not or not you utilize it in any respect. That requires buying a radical understanding of those instruments’ talents and pitfalls, being meticulous concerning the parameters of your prompts, and a willingness to adapt your course of regularly. It’s an ongoing course of, and it wants guardrails—equivalent to “at all times” and “by no means” instructions to keep away from particular issues and (human) fact-checking. In any other case, you’re taking part in with a gun that might simply go off.
There are systemic safeguards past easy methods. Whether or not you’re an unbiased author or a full newsroom, it pays to have an AI coverage. As a media AI trainer, I after all would encourage investing in coaching, however I believe it’s nonetheless objectively a good suggestion. However most significantly, the trial-and-error that comes with determining the boundaries of “good AI” must be stored out of public view should you can keep away from it. Within the case of AI-assisted writing, creating your prompting and guardrails in a non-public sandbox is important.
That will appear apparent, however a part of the “magic” of AI is that it creates outputs that appear similar to human-created outputs which have gone by way of a rigorous course of. To the untrained eye, the looks of competence feels adequate. Unlocking AI’s potential as a associate in writing and journalism means not merely trusting the underlying course of, however accepting your position to construct it, take a look at it, and alter it as wanted. The extra journalists do this, the extra the stigma will fade.
{“blockType”:”mv-promo-block”,”information”:{“imageDesktopUrl”:”https://photos.fastcompany.com/picture/add/f_webp,q_auto,c_fit/wp-cms-2/2025/03/media-copilot.png”,”imageMobileUrl”:”https://photos.fastcompany.com/picture/add/f_webp,q_auto,c_fit/wp-cms-2/2025/03/fe289316-bc4f-44ef-96bf-148b3d8578c1_1440x1440.png”,”eyebrow”:””,”headline”:”u003Cstrongu003ESubscribe to The Media Copilotu003C/strongu003E”,”dek”:”Need extra about how AI is altering media? By no means miss an replace from Pete Pachal by signing up for The Media Copilot. To study extra go to u003Ca href=u0022https://mediacopilot.substack.com/u0022u003Emediacopilot.substack.comu003C/au003E”,”subhed”:””,”description”:””,”ctaText”:”SIGN UP”,”ctaUrl”:”https://mediacopilot.substack.com/”,”theme”:{“bg”:”#f5f5f5″,”textual content”:”#000000″,”eyebrow”:”#9aa2aa”,”subhed”:”#ffffff”,”buttonBg”:”#000000″,”buttonHoverBg”:”#3b3f46″,”buttonText”:”#ffffff”},”imageDesktopId”:91453847,”imageMobileId”:91453848,”shareable”:false,”slug”:””,”wpCssClasses”:””}}
