All articles
AI Wrote This Headline: On Algorithmic Editing and the Loss of Voice
As publications adopt AI editing tools, something subtle is changing — not just how content is produced, but how it sounds. And readers are noticing.
The headline on this story was written by a person. The three alternative headlines below were generated by an algorithm trained on the publication's archive:\n\n1. "AI-Generated Headlines: What Editors Need to Know"\n2. "How AI Tools Are Reshaping Newsrooms"\n3. "Journalism in the Age of Automation: Key Insights"\n\nNone of them are wrong. None of them are mine.\n\nI ran this test at three publications that use algorithmic headline tools. At each one, the AI-generated options were accurate, SEO-optimized, and completely devoid of personality.\n\n## The stealth homogenization of newsroom voice\n\nEvery publication has a voice — a set of editorial instincts that shape which stories get told, from what angle, with what emphasis. That voice is developed over years, through editorial culture, hiring decisions, and institutional memory.\n\nAlgorithmic tools don't have voice. They have optimization targets: CTR, time-on-page, social shares. These metrics are legible, measurable, and almost completely disconnected from the thing that makes journalism valuable: trust.\n\nA headline optimized for clicks might be sensational. A headline optimized for shares might be vague. A headline optimized for time-on-page might bury the lede.\n\nThe publication's voice becomes a function of the algorithm's preferences, not the editor's judgment.\n\n## What readers actually reported\n\nIn exit interviews with former subscribers who cited "generic content" as their reason for leaving, several mentioned headline fatigue — reading what felt like the same headline in different publications, not realizing the AI was trained on the same industry benchmarks.\n\nOne reader put it simply: "I stopped feeling like anyone was talking to me."\n\n## The practical argument for keeping humans in the headline room\n\nHeadlines aren't just labels. They're the first editorial decision — what the story is about, what angle to take, what tone to set. Making that decision requires knowing what you believe.\n\nAlgorithmic tools can optimize for performance. They cannot evaluate whether a story is worth telling, whether the angle serves the reader, or whether the publication's credibility is being used responsibly.\n\nThe editor who writes the headline is doing more than SEO work. They're exercising editorial judgment — the thing that makes a publication worth subscribing to in the first place.