Search feature returns chatbot-style responses for terms like disregard, ignore, dismiss; company acknowledges issue, promises fix, advises using standard results.
Generative AI tools are once again in the limelight, after a simple search for the term “disregard” in Google’s AI Overviews triggered what looked like a chatbot-style command response instead of a dictionary definition.
The issue appears when users enter words such as “disregard”, “ignore”, “dismiss”, “stop”, or even “remember”, and the system responds as though it is following a prompt rather than interpreting a query. Instead of delivering a definition, the tool has been shown responding with messages such as “Understood. Message disregarded,” followed by a large blank area before the normal search results appear, as described in a Mac Rumors report.
The glitch has become the latest example of how Google’s AI search features can mistake ordinary language for instructions, a problem that surfaces just as the firm pushes further into an AI-first search paradigm at its I/O event last week.
Google has acknowledged the bug and is working on a fix, while stressing that the problem is tied to AI Overviews rather than the broader search changes unveiled at the event. The firm said the system is misreading some action-related queries and that an update should roll out soon.
According to an Android Authority report, this vulnerability fits a familiar pattern for AI Overviews, which has drawn criticism before, for generating confident but faulty outputs. Earlier problems had included incorrect or unsafe answers on basic topics, reinforcing concerns that the feature still struggles to separate user intent from text it is meant to process.
In the meantime, the safest workaround for users is to rely on standard search results below the AI panel, or use more specific wording that does not resemble a command. The episode is a reminder that AI-assisted search can still blur the line between content and instruction in ways that ordinary users can notice immediately.
