Particle.news

Grok Chatbot Sent Gibberish to Some Web Users

A temporary generation glitch that users say clears with a fresh chat exposes reliability concerns for Grok's service and growth prospects.

Overview

  • Users reported that Grok delivered nonsensical, multi-paragraph “word salad” replies starting Wednesday while using Grok Lite and direct queries on Grok.com.
  • The official Grok account on X called the problem a rare temporary generation glitch and advised starting a fresh chat or regenerating responses, while the x.ai status page showed no incidents.
  • TechCrunch and other outlets found the issue visible on user forums but could not reproduce it in their own tests, indicating the fault likely affects a limited subset of web users.
  • Affected users flooded the Grok subreddit with complaints and shared examples that sometimes included odd source links, and some people said refreshing fixed the output while others continued to see gibberish.
  • The glitch deepens worries about Grok’s reliability after recent staff departures and a July Opus-class model release and comes as polls and analytics show low user satisfaction and tiny paid-conversion rates, which could magnify reputational and commercial damage.