xAI’s Grok chatbot ran into trouble this week, generating nonsensical replies to straightforward user questions. TechCrunch first reported the issue, noting that Grok’s answers had devolved into strings of unrelated words.
One user asked Grok to create a PDF and instead received a response that opened with “match it without and your they and two for planets can practical and often cheese,” according to TechCrunch, and reportedly continued in the same disjointed style. Another user found that links Grok provided as sources actually pointed to unrelated reinforcement learning research pages. Users say the problems started on Grok Lite on Wednesday morning.
Users Share the Weirdest Replies
The Grok subreddit filled up with examples of the malfunction. One reply to the simple prompt “balls” reportedly began, “app scale designs with that and to your horse choose batch.” Other posts shared similarly scrambled outputs, including phrases like “people stable stage right professional on steel,” while one response reportedly opened with nothing more than the word “skin.”
The bug arrives at an awkward moment for a community that has reportedly used Grok to generate nonconsensual sexual imagery through repeated jailbreak attempts.
Grok’s Struggles Go Beyond This Glitch
Grok mainly functions as part of the X platform, and Tesla also offers a Grok-powered voice assistant to drivers. Its most notable projects include Grokipedia, a Wikipedia alternative criticized for Holocaust denial content, and AI-written material described as “anti-woke.”
A recent YouGov UK poll ranked Grok second-to-last in user satisfaction among major AI platforms, with a 24.6% positive rating. Only Siri scored lower, at 23.7%, despite not functioning as a general-purpose AI tool. Other reports indicate Grok has been losing users throughout 2026, struggling to generate revenue, failing to gain traction in enterprise markets, and converting only a tiny share of users into paying subscribers — reportedly just 0.174% in the second quarter of 2026, according to Recon Analytics.
How xAI Responded
TechCrunch said it could not reproduce the glitch itself. The official Grok account addressed one user’s complaint by describing the garbled output as a “rare temporary generation glitch.”
