Google Search caught rewriting news headlines with AI

TL;DR:

• Google has confirmed it is running a “small” experiment replacing publisher-written headlines in traditional search results with AI-generated alternatives — not just in Google Discover, where it already does this.
• At least one rewritten headline appeared to endorse a product the original article criticised, showing how meaning can shift when editorial intent is removed.
• The test follows Google’s pattern of first calling features “experiments” before rolling them out permanently, as it did with Discover’s AI headlines earlier this year.

Google’s search results have long been treated as a relatively faithful index of the web. Publishers write headlines, Google displays them, and users click through. That compact is now fraying. The Verge reports that Google is replacing headlines written by its journalists — and those of other publications — with AI-generated versions in the traditional “10 blue links” results.

In one example, a Verge headline reading “I used the ‘cheat on everything’ AI tool and it didn’t help me cheat on anything” was reduced to five words: “‘Cheat on everything’ AI tool.” The rewrite stripped the critical angle entirely, making it sound like a product recommendation.

Google’s position

Three Google spokespeople confirmed the test but described it as “small” and “narrow,” with no approval for a wider launch. They said the aim is to “identify content on a page that would be a useful and relevant title to a user’s query” and that the experiment applies across websites, not just news. Google claimed any eventual production version would not use generative AI, though it did not explain what alternative technology would replace publisher headlines without it.

The Verge’s editor noted that Google has long made minor adjustments to how headlines appear — truncating long titles or swapping in on-page headlines for search-specific ones. But generating entirely new text is a different matter, one he says he has not seen in 15 years of editing.

A familiar playbook

There is precedent for concern. Google initially described its AI-generated headlines in Google Discover as an experiment. Within a month, the company reclassified the feature as permanent, saying it “performs well for user satisfaction.” The Verge has documented cases where Discover headlines misrepresented stories — including one that reversed the meaning of a foreign policy report.

For UK publishers already contending with declining referral traffic and the growing dominance of AI Overviews, this adds another layer of risk. If Google can alter what a story appears to say before anyone clicks, publishers lose control of their editorial voice at the point where it matters most.

Looking forward

This test may be small today, but the direction of travel is clear. Google is steadily inserting itself between publishers and their audiences, not just by summarising content in AI Overviews but now by rewriting the labels on the content itself. For news organisations, the question is no longer whether Google will change your headline, but when.