Google Search now changes the headlines of AI-generated news, without fear of distorting meanings

Google Begins Replacing News Headlines with AI-Generated Versions

The search engine Google has started replacing original article headlines in its results with versions generated by artificial intelligence (AI). Previously, this approach was used only in the Discover feed; now it also affects traditional search results.

How It Manifests

*The Verge* discovered several instances where Google replaced authorial headlines with its own. Sometimes the new phrasing distorted the original meaning. For example, an article about “testing an AI tool for cheating across everything that didn’t meet a student’s expectations” was shortened to “AI Tool for Cheating Across Everything.” Such a change could give readers a false impression that the outlet is recommending the tool itself.

Google’s Response

Company representatives—Jennifer Kutz, Mallory De Leon, and Ned Adrians—stated that this is a “small” experiment. They did not specify the scale of testing but noted that changes affect not only news sites but other websites as well. The goal of the experiment is to determine content that could become a useful headline responding to user queries.

Journalists’ Concerns

Despite official reassurances, *The Verge* expresses concern that the experiment could become a permanent feature. Earlier, Google said AI headlines in Discover were just a test and later announced them as a “feature that increases user engagement.” The company did not explain why it refuses to follow headline rules it had recommended to publishers.

Difference from Previous Changes

Until now, Google made only minor edits: trimming headline length or slightly adjusting wording. Now the algorithm creates text from scratch, ignoring journalists’ efforts to craft honest and engaging titles without clickbait. Journalists compared the situation to a bookstore that removes book covers and changes their titles.

Current State

At present, altered headlines are rare and do not contain as severe errors as in Discover (e.g., incorrect information about PlayStation Portal or lifting drone bans). Nevertheless, outlets fear that the test is a precursor to larger-scale changes.