The debate over AI in journalism just got personal. WIRED's Steven Levy—a four-decade veteran who's chronicled everything from the birth of personal computing to the rise of Google—is pushing back hard against AI-assisted writing tools creeping into newsrooms. In a sharp column published today, Levy argues the efficiency gains publishers are chasing may come at a cost the industry hasn't fully reckoned with: the erosion of the craft that makes journalism worth reading in the first place.
Steven Levy isn't mincing words. The WIRED editor-at-large and author of seminal tech books like "Hackers" and "In the Plex" just drew a hard line in the sand: AI can draft a lot of things, but it's not writing his stories.
His column, published this afternoon under the headline "AI Drafting My Stories? Over My Dead Body," tackles what's become an uncomfortable question in newsrooms across the industry. As publishers face relentless pressure to do more with less, AI writing assistants have started appearing in editorial workflows—sometimes with writer buy-in, sometimes without it. Levy's argument cuts through the productivity rhetoric to something more fundamental: what gets lost when machines do the writing?
The timing is pointed. Major news organizations have been quietly testing generative AI tools for months. The Associated Press has used automation for earnings reports since 2014, but today's large language models promise—or threaten, depending on your perspective—to handle far more complex storytelling. Forbes experimented with AI-generated draft articles earlier this year. CNET faced backlash in 2023 after readers discovered AI-written explainers riddled with errors, forcing editors to issue corrections and pause the program.
Levy's concerns go beyond accuracy. "AI-assisted writing is creeping into newsrooms under the guise of efficiency," he writes, according to the column's summary. "But the tradeoff may be more profound than publishers are willing to admit."
That tradeoff involves the invisible labor that makes journalism journalism—the judgment calls about what details matter, the cultivation of sources who trust you enough to go on record, the narrative instincts that turn a collection of facts into a story readers actually care about. These aren't inefficiencies to optimize away. They're the point.
The essay lands as the AI industry itself is grappling with questions about synthetic content. OpenAI has positioned its GPT-4 and upcoming models as writing assistants, not replacements, but the distinction gets blurry fast when publishers are measuring cost-per-article. Microsoft, which has integrated AI writing tools into everything from Word to Edge, is actively courting media companies with enterprise solutions.
Levy has spent his career documenting tech revolutions, often as an early enthusiast. He embedded with Google for his 2011 book "In the Plex," gaining unprecedented access to the company's inner workings. He chronicled the hacker ethos that birthed Silicon Valley. His skepticism about AI in newsrooms carries weight precisely because it's not coming from a Luddite—it's coming from someone who understands both the technology and what's at stake.
The editorial arrives amid broader anxiety about AI's impact on creative work. Hollywood writers struck partly over AI contract language last year. Artists have sued companies like Stability AI over training data scraped without permission. But journalism faces a unique pressure: the business model was already broken before AI showed up offering to make content production cheaper.
That's the trap Levy seems to be highlighting. When your industry is hemorrhaging revenue and AI vendors promise to cut costs, the temptation to automate is enormous. But if the result is journalism that readers can't distinguish from the slop already flooding the internet, what exactly are publishers saving?
The column doesn't offer easy answers, which is probably the point. This isn't a problem with a clean technical solution. It's a question about what journalism is for—whether it's a commodity to be produced as cheaply as possible, or a craft that requires human judgment, experience, and yeah, inefficiency.
For now, at least one prominent byline won't be attached to AI-generated drafts. Whether that stance becomes a rallying cry or a rearguard action depends on how the rest of the industry responds.
Levy's essay crystallizes a tension the tech industry has been dancing around: efficiency isn't always the right metric. Sometimes the slow, expensive, deeply human parts of a process are what make it valuable. As publishers continue experimenting with AI tools, the question isn't just whether machines can write news—it's whether anyone will care to read it if they do. For journalists watching their profession get squeezed from every direction, Levy's line in the sand might be exactly the provocation the industry needs.