We are currently beta-testing a new AI tool for fact-checking at Refly, but I’ve found that it cannot replace the nuance of a human eye when it comes to voice, tone, and argument.
Last month, I asked our team to review a draft of a piece on climate policy. The piece was well-sourced, the data was sound, and the grammar was impeccable. It was, by any machine metric, a "perfect" article. Yet, when I read it aloud to myself, it felt flat. It lacked the friction that makes an argument compelling.
That is the crux of the debate we are having in newsrooms right now. AI can catch a misspelled name or a citation error in seconds. It can generate a first draft of a press release in minutes. But it cannot feel the weight of a sentence. It cannot know that a word choice feels "off" not because it is grammatically incorrect, but because it clashes with the writer's established rhythm.
"The computer can spot the error, but only the editor can understand the argument."
— Elena Vance, Senior Editor
At Refly, we are leaning into this distinction. Our editors spend more time on the "second draft" than the first. We don't use AI to write for us; we use it to handle the logistical heavy lifting—checking dates, formatting references, and flagging passive voice so that our human editors can focus entirely on structure, clarity, and resonance.
This is why the "Craft" section of our magazine matters. We publish essays not just to inform, but to model the art of thinking. Our editorial guidelines specifically ask writers to embrace complexity, to prioritize nuance over brevity, and to trust their reader's intelligence. An AI can summarize a complex topic; only a human can interrogate it.
So, what does a human editor do? They listen. They ask questions. They make the difficult choices about what to cut and what to keep, guided not by a set of algorithms, but by a deep, human understanding of what makes a story matter.
