Overview
- Google confirmed to The Verge it is running a small UI experiment for a subset of Discover users to make topics easier to digest.
- Observed AI headlines are uniformly shortened to four words, with documented misfires such as "BG3 players exploit children," "Steam Machine price revealed," and "AMD GPU tops Nvidia."
- The AI provenance is not clearly surfaced to readers, with a warning only visible after tapping "More," raising the risk that publishers are blamed for the wording.
- Publishers and journalists caution the auto-generated titles can damage trust and reduce referral traffic when headlines distort or oversimplify coverage.
- Analysts note the test appears to conflict with Discover rules requiring representative, non-clickbait titles, and Google has not detailed rollout plans or publisher opt-out options.