Grok Is Sending Users Word Salad Instead of Answers
A bug in xAI's Grok chatbot is filling some users' screens with meaningless strings of words. Here is what is happening, who is affected, and what to do if it hits you.
Key points
- Grok Lite, the lighter version of xAI's AI chatbot, began producing gibberish responses for some users as early as Wednesday morning, August 20, 2026.
- One affected user asked Grok to create a PDF and received: "match it without and your they and two for planets can practical and often cheese."
- The Grok account on X confirmed the bug Thursday morning, calling it "a rare temporary generation glitch."
- xAI said its status page showed all services operational, and suggested starting a fresh chat to clear the problem.
- xAI has lost at least 50 researchers and engineers in recent months, according to a May 2026 report.
Something broke inside Grok this week, and a lot of users noticed before xAI said a word about it.
Affected users of Grok Lite, the lower-powered tier of xAI's AI chatbot, began posting screenshots of bizarre responses on Wednesday. One person asked the chatbot to generate a PDF document and got back a string of disconnected words: "match it without and your they and two for planets can practical and often cheese," followed by several more paragraphs of the same. Another user found their response packed with links to reinforcement learning research sites. Reinforcement learning is a training method where an AI system learns by trial and error, similar to how you might teach a dog with rewards.
The Grok subreddit, a community forum for Grok users on Reddit, filled up with complaints. TechCrunch first reported the issue after speaking with affected users.
Is this serious?
For now, the bug looks contained. It appears limited to direct conversations on Grok.com and has not touched the Grok account on X, which kept posting normally. xAI's own status page listed all services as operational throughout.
The Grok team confirmed the problem on X Thursday morning: "That pure word salad is a rare temporary generation glitch. Start a fresh chat or regenerate, it usually clears right away. Sorry about the gibberish."
Refreshing the chat session fixed the problem for many users. Some reported the nonsense kept coming even after several refreshes, though it is unclear how many people hit that wall. TechCrunch could not reproduce the issue in its own testing, which suggests only a slice of users were affected.
xAI did not respond to a request for comment on what caused the glitch.
What does this mean for Grok users?
If you see random word strings in a Grok response, start a fresh chat. That single step clears it for most people. If the problem persists, check status.x.ai for any official updates.
This glitch arrives at a tricky moment for the company. xAI lost most of its founding team and at least 50 researchers and engineers in recent months, a significant number for a young AI lab. The company released a new foundation model in July 2026, describing it as faster and cheaper than the previous generation.
The honest takeaway: AI chatbots can produce strange output when something goes wrong on the server side. It is not your device, it is not your account, and it is usually not permanent. Start a fresh chat, and if the problem sticks, wait an hour and try again. No important decisions should rest on a single chatbot response anyway.
