Reports of the malfunction emerged among users of Grok Lite, with some saying ordinary requests produced long strings of unrelated words or irrelevant links. TechCrunch was unable to reproduce the fault, suggesting it affected only part of the user base.

The problem initially appeared concentrated on direct conversations through Grok's website rather than the Grok account operating on X.

xAI's Grok account described the problem as a temporary generation glitch and advised affected users to start a new conversation or regenerate their response. Some users said refreshing fixed the problem, while others reported continued failures.

At the same time, xAI's public service-status system continued to list Grok's web, mobile and X services as available, with no active incident declared.

The company has not publicly identified the underlying technical cause of the malformed responses.

The episode illustrates a broader reliability problem facing generative AI providers: a service can remain technically online while still producing output that is unusable for some customers.