Google is testing artificial intelligence to generate news headlines

Google continues to expand the use of artificial intelligence in its services, and it doesn’t work very well. Journalists from The Verge noticed that in Google Discover, the company is testing replacing article titles with generated versions – sometimes inaccurate or even erroneous.

Some of these replacements have been misleading to users. For example, Discover showed an Ars Technica article under the title “Steam Machine price revealed”, although the article only talked about the design of the new console platform, and Valve did not report anything about the price. Engadget also drew attention to another format of the experiment – short AI summaries appeared alongside the original titles.

Google adds a “Generated with AI, which can make mistakes” label under such blocks, but this does not remove criticism – publishers and users point out the risks of misinformation and distortion of the content of the news.

Company spokesperson Mallory DeLeon explained that this is a “small interface experiment” for a portion of the Discover audience. According to her, the test is aimed at changing the presentation of headlines to make it easier to get acquainted with the topic before moving on to the material.

Despite these explanations, the situation is causing concern among online media, which have already been in conflict with Google for years over the use of their content in search and news services. Publishers have repeatedly demanded compensation, and Google has sometimes responded by limiting access to their materials in search results.

In parallel, the company continues to integrate AI into its main search. Google Search VP Robbie Stein confirmed that Google is testing a combination of AI Mode and AI Overviews on the same screen in the mobile interface – thus, AI suggestions will become even more noticeable in search results.