Grok keeps sending gibberish responses to users
An xAI generation glitch is turning Grok into a word-salad machine for some users — and the company's own status page still insists nothing is wrong.
xAI's Grok chatbot has been replying with pages of incoherent gibberish to a subset of users, with the failures concentrated in the free Grok Lite tier and surfacing as early as Wednesday morning. One user who asked Grok to generate a PDF got back "match it without and your they and two for planets can practical and often cheese…" with the nonsense continuing for several paragraphs; another, checking the model's cited links, found a string of references to reinforcement-learning research papers. TechCrunch, which first reported the issue, was unable to reproduce it and says the problem appears limited to direct queries on Grok.com, with the Grok account on X.com unaffected — so it looks like a backend generation bug rather than a prompt or app-layer failure.
xAI has not commented on the incident, and the official status page still shows all Grok services "fully operational with no incidents." The Grok account did confirm the problem on X Thursday morning, calling it "a rare temporary generation glitch" and advising users to start a fresh chat or regenerate, which "usually clears right away." That matches what affected users are reporting: refreshing the session often restores normal output, though some said the gibberish persisted across multiple refreshes. The Grok subreddit has been flooded with complaints, a useful signal that the failure, while narrow, is persistent enough to break the product experience for the people hitting it.
Why it matters: glitches like this are the unglamorous cost of shipping frontier models fast, and they land at an awkward moment for xAI. The company released Grok 4.5 in July, pitching it as an "Opus-class model, but faster, more token-efficient and lower cost," and has been bleeding senior talent — TechCrunch notes it has lost most of its founding team and at least 50 researchers and engineers since its merger with SpaceX. A model that occasionally forgets how to form sentences undermines the reliability story a lab needs when it is competing on being production-grade, not just benchmark-leading. It is also a reminder that "status: all systems operational" is a claim a company makes, not a guarantee a user can rely on.
What to watch: whether xAI acknowledges the root cause, and whether the gibberish is tied to the Lite tier specifically or to a shared generation path.
If your AI assistant started replying in word salad, would you trust a fresh chat — or start looking for a rival? Tell us in the comments.
