Google is testing an AI feature inside its Discover news feed that changes how some news headlines appear to users. The test automatically rewrites editorial headlines, sometimes making them shorter, more provocative, or incorrect.
The issue is not only that Google is experimenting with AI summaries. The concern is that the rewritten headlines can shift the meaning of a newsroom’s original framing while appearing in a product where Google already controls how news is delivered.
What Google Is Testing In Discover
The feature appears in Google’s "Discover" news feed, where users encounter article recommendations. According to the source, the AI system automatically rewrites editorial headlines rather than simply displaying the wording chosen by the publication.
Google told The Verge that the feature is a small test. The company said the goal is to help users grasp articles more quickly.
That explanation points to a real product problem: headlines need to communicate fast. But the examples in the source show why speed can become a problem when an automated rewrite removes context or states something the original headline did not say.
Why The Rewrites Are Drawing Criticism
Headlines are not just labels. They are editorial choices. A newsroom decides what to emphasize, what to qualify, and what not to claim. When an AI system replaces that wording, it also replaces part of the editorial judgment behind it.
The source describes the rewritten versions as the kind of engagement-focused headline Google warns against in its own Discover rules. Those rules explicitly reject clickbait.
That creates the central tension in this test. Google’s own policy discourages clickbait, yet the AI-generated versions cited in the source can make headlines shorter, sharper, and more provocative. In one case, the rewrite was not merely punchier; it was inaccurate.
The Examples Show The Risk
One example came from Ars Technica. Its original headline was: "Valve's Steam Machine looks like a console, but don't expect it to be priced like one". In Google Discover, that became: "Steam Machine price revealed."
The difference matters. The original headline signaled that the article was about expectations around pricing. The rewritten version suggested that a price had been revealed. According to the source, that was inaccurate.
A second example involved a Mindfactory report. The original headline was: "Radeon RX 9070 XT Outsells The Entire NVIDIA RTX 50 Series On Popular German Retailer". Google’s rewrite shortened it to: "AMD GPU tops Nvidia."
That version is easier to scan, but it is also much broader. The original headline included the product name, the comparison, and the setting: a popular German retailer. The rewrite reduced those details into a general claim about AMD and Nvidia.
What Changes When AI Reframes News
The practical effect of a headline rewrite happens before a reader reaches the article. A user may decide whether to open a story based on Google’s version, not the newsroom’s version.
That is why even small wording changes can matter. A shorter headline can make an article feel more definitive than it is. A provocative headline can change the reader’s expectation. An inaccurate headline can mislead before the article has a chance to explain itself.
The source also raises a broader concern: Google’s automated framing may tighten the company’s grip on how news is shaped and delivered. That concern builds on Google’s existing influence over distribution. If Google not only chooses what appears in Discover but also rewrites how it is presented, the company’s role expands from delivery into editorial framing.
The Open Question For Discover
Google’s stated aim is faster comprehension. The examples show the tradeoff: making a headline easier to grasp can also strip away the details that keep it accurate.
For newsrooms, the issue is control over presentation. For readers, it is whether the headline they see reflects the article they are about to read. For Google, the test raises a simple question: can an AI rewrite headlines without producing the kind of clickbait-style framing its own Discover rules reject?
Based on the examples in the source, that question remains unresolved. The test may be small, but it touches a large issue in online news: who gets to decide how a story is framed at the moment readers first encounter it.