AI Shrank the Newsroom. Now Tech Companies Are Competing for Journalists.

U.S. journalism has lost nearly half its editorial jobs since 2008. In the same period, Meta, Google DeepMind, and OpenAI have been actively recruiting former reporters, producers, and editors. The new job titles: "Content Engineer," "Model Designer," "AI Philosopher." Why do AI companies want journalists? For a decade, the economics of journalism have been quietly collapsing. You write articles, search engines bring readers, readers generate ad revenue. Generative AI broke that loop. AI platforms now give users answers directly, so readers no longer click through to the original article. Without traffic, the commercial case for investigative reporting and long-term coverage slowly disappears. But the skills journalists spent twenty years developing didn't disappear with it. They know how to read what someone is actually asking, not just what they said. They know when to push and when to stop, how to adjust the same information for different audiences, where the cultural landmines are. These instincts don't come from a handbook. They accumulate through enough conversations, enough stories, enough corrections. Bianca was a director at VICE and a senior producer at NBC News before joining Meta to work on AI. She laid out the problem with a clear example: if someone asks whether Taylor Swift will get married at Madison Square Garden, the worst AI response is a fabricated answer. The second worst is an information dump. A good response honestly flags the speculation while reading what the person actually wants to know, not just the literal question. AI still can't quite do this on its own. Tony spent six years hosting a podcast. One year he got sick and couldn't record, so he built a voice agent to host in his place. Strict system prompts didn't hold. A multi-agent setup collapsed under too many constraints. Eventually he organized six years of podcast archives and had the system pull relevant examples into each conversation before it started. That worked. Google later recruited him to DeepMind. His resume said "Podcast Host." What Google wanted was six years of instinct for which questions are worth following. Tony is now a Model Designer at Google DeepMind. American journalism is missing someone who spent six years doing that.