Google is testing AI‑written titles in Search

Google Search is running a new test that replaces normal page titles and headlines with AI‑generated ones in the search results. Instead of showing the title that the publisher wrote, Google’s systems can now show a different title that they think fits the user’s search better.

Publishers first noticed this when their articles started to appear in Google with new wording they had not written. In some cases, the meaning of the AI‑written title was different from the original, or important details were removed.

How this test works

Google says this is a “small” and “narrow” experiment. The system looks beyond the normal HTML title tag and may also read headings and other text on the page to build a new title for the search result. The goal, according to Google, is to match the searcher’s intent better and help people understand why a result is relevant.

Google also confirmed that this test uses generative AI today, but said that if they launch a full feature later, it would not be based on a generative model in the same way. The company did not give clear technical details on how that future version would work.

This started in Google Discover

This is not the first time Google has rewritten headlines with AI. In late 2025, Google Discover – the content feed on Android phones and inside the Google app – began showing AI‑generated titles instead of the original ones from publishers. Reports showed several bad examples, including headlines that were misleading, too clickbaity, or simply wrong compared to the article.

At first, Google also called that use of AI headlines in Discover an “experiment”. Later, it described them as a product feature, which made some people worry that the same thing would move into main Google Search – which is exactly what is now being tested.

Why publishers are worried

Publishers and site owners are concerned for several reasons:

• They can lose control over how their work is framed and presented to readers.

• AI‑written titles can change the meaning, add exaggeration, or even introduce facts that are not in the article.

• If the AI title is confusing or wrong, users may blame the publisher, not Google.

• Traffic might fall if a new title attracts fewer clicks or targets the wrong audience.

Some headlines in earlier tests have already shown this risk, like game or tech stories being turned into more extreme or inaccurate titles.

Google’s response and next steps

Google says it is trying to improve how AI‑driven search presents sources and promises better ways for people to check where information comes from. The company has also admitted that some AI‑generated summaries and features have produced strange or incorrect results in the past and says it is limiting when such summaries appear and working to make them more reliable.

For now, the AI‑title test is limited, and Google has not announced a full launch or tools for publishers to opt out. But if this experiment expands, it could change a basic part of the web: who really writes the headline that most users see – the publisher, or Google’s AI.