Google used to promise a simple deal: you search, you see headlines written by publishers, you click. Now the company is rewriting those headlines with AI before you ever see them.
What started as a "small UI experiment for a subset of Discover users" in late 2025 has expanded into Google Search proper. And Google is no longer calling it a test. A spokesperson told The Verge the AI-generated headlines are a feature that "performs well for user satisfaction."
Four Words, Zero Accuracy
The AI appears obsessed with compressing complex stories into roughly four words, and the results range from misleading to outright false. A PC Gamer article about players discovering how to clone virtual children in Baldur's Gate 3 to break the game became "BG3 players exploit children." An Ars Technica piece whose headline explicitly said Valve's Steam Machine won't be priced like a console was rewritten as "Steam Machine price revealed." A Verge story by Tom Warren about how Microsoft developers use AI had the word "How" stripped from the title, changing the meaning entirely.
Google does include a small disclaimer: "Generated with AI, which can make mistakes." But users only see it if they tap "See more," which means most people scrolling through search results have no idea the headline they are reading was never written by the publication it links to.
Publishers Lose Control, Then Lose Traffic
This matters because headlines are not just labels. They are editorial decisions. A headline frames what a story is about, sets expectations, and represents the publication's judgment. When Google rewrites them, it severs the connection between what a journalist intended to communicate and what millions of readers actually see.
The timing makes this worse. Small publishers are already losing up to 60% of their Google referral traffic according to a March 2026 study, largely because AI Overviews answer queries directly without sending users to source websites. Breaking news traffic is up 103% across Google surfaces since late 2024 - not because publishers are thriving, but because breaking news is the one category where AI still struggles to replace human reporting in real time.
So the picture is: Google is simultaneously taking publishers' traffic with AI Overviews and rewriting their remaining headlines with AI. Publishers get fewer clicks, and the clicks they do get come from headlines they did not write and cannot control.
The "User Satisfaction" Problem
Google's defense - that AI headlines "perform well for user satisfaction" - reveals the core tension. Satisfaction for whom? Users might prefer shorter, punchier titles in a feed. But if those titles are inaccurate, users are being satisfied by misinformation. And publishers, whose content Google depends on entirely, were never asked.
This is a fundamentally different move from AI Overviews, which at least attempted to synthesize information. Rewriting someone else's headline and displaying it under their name is closer to putting words in their mouth. The fact that Google tested this quietly, then promoted it to a permanent feature without any public discussion with the news industry, tells you everything about how the company views the relationship.
For anyone who publishes content online - bloggers, news outlets, businesses with editorial content - this is worth paying close attention to. Your headline is no longer necessarily your headline once Google indexes it.