For more than two decades Google Search offered a simple promise: you click a result and you get the publisher’s page as the publisher presented it. That tidy arrangement is fraying.

What is happening

Google has begun replacing the headlines publishers write with AI-generated alternatives inside standard Search results. This is not just limited to the Discover feed anymore. Staff at a major tech publication found multiple instances where the headline shown in Search did not match the original headline on the page and sometimes altered its meaning.

Notable examples

  • One long-form headline that read: I used the ‘cheat on everything’ AI tool and it didn’t help me cheat on anything was shortened in Search to: 'Cheat on everything' AI tool. That trim makes the headline sound like an endorsement instead of a report.
  • Another piece about a company update ended up as a bland new headline: Copilot Changes: Marketing Teams at it Again. That version did not follow the publisher’s style and was created by Google’s system.

What Google says

Google describes the work as a small, narrow experiment and confirmed it is testing generative AI to identify content on pages that could serve as a useful title for a user’s query. Three Google spokespeople named as contacts for this explanation were Jennifer Kutz, Mallory De Leon, and Ned Adriance. They would not provide details about how many users or queries are involved.

Google also said the test is not specific to news and is intended to improve how titles match user queries across the web. At the same time the company stated that if the experiment ever becomes a full product, it would not create headlines with generative AI. That leaves an open question about how Google might replace publisher titles without generative models.

How this differs from past behavior

Search engines have long adjusted how they display a site’s title. Typical edits include trimming long headlines or showing an alternative headline already present on the page, such as the on-page headline when a separate search headline field exists. Editors have planned around those behaviors.

What is new here is that Google’s system appears to be composing entirely new headlines in some cases, rather than only selecting or trimming existing text. For newsrooms that invest time in precise, carefully worded headlines, that is a meaningful shift.

Why publishers and readers should care

  • Meaning can change. A shortened or rewritten headline can give users a different impression of a story before they click.
  • Trust is at stake. When institutions alter how journalism is presented, readers may find it harder to rely on headlines as accurate pointers to the underlying article.
  • It may not stay an experiment. Google previously labeled an AI headline change in Discover as experimental before making it a feature that the company said performed well for user satisfaction. That history makes the current test consequential.

Context and disclosure

Some headlines produced by Google’s AI in other places have been inaccurate. The changes in standard Search results appear less extreme so far, but they may be an early sign of broader change.

For full transparency, the publisher affected by these examples is part of a larger company that has filed a lawsuit against Google alleging unlawful ad tech practices.

Bottom line

Google is experimenting with generating new headlines for pages shown in Search. The company says the intent is better alignment with user queries, but it has not explained the boundaries of the test or how publishers will be informed. For newsrooms, the shift raises clear editorial and credibility questions. For readers, it means the headline you see in Search may not be the headline the publisher wrote.