If ChatGPT won’t open, can’t send messages, or keeps spinning, it’s usually not that “the system is broken,” but rather a combination of network issues, browser cache, extension blocking, or usage limits. The following ChatGPT troubleshooting steps go from “quick first, detailed later,” narrowing the problem down step by step into something you can handle. Each step can be verified independently, and after completing them you can generally pinpoint the cause.
Start with three quick checks: confirm whether it’s on your side or the server side
When doing ChatGPT troubleshooting, first switch to another network or directly switch to a mobile hotspot and see if it recovers immediately; if it does, suspect the local network first or blocking by a workplace/campus network. Second, open status.openai.com to see whether there are service disruptions; if the server side is having issues, there’s no need to keep tinkering with your device. Third, log in once using an incognito/private window—this quickly rules out cache, cookies, and extension conflicts, and is the most time-saving ChatGPT troubleshooting method.
Blank page, infinite loading: most likely cache or scripts being blocked
For this type of ChatGPT troubleshooting, start with the browser: first do a hard refresh (Windows: Ctrl+F5; Mac: Cmd+Shift+R), then clear site data (clear only cookies and cache related to chatgpt.com / openai.com). If you have ad blockers, script managers, or privacy/anti-tracking extensions installed, temporarily disable them and test again, because they may be blocking required scripts or WebSocket connections. Finally, test with a different browser (switch among Chrome/Edge/Firefox) to quickly determine whether it’s a browser-environment issue.
Message send failures and 401/403: troubleshoot login state and account risk controls
When doing ChatGPT troubleshooting, if the page opens but messages fail to send, log out and log back in first—many issues come from an expired login state or token confusion caused by switching across devices. A 401 usually means unauthorized access or an expired session; re-logging in and clearing site cookies often works. A 403 is more like access being blocked or a security policy being triggered; it’s recommended to disable proxy split-tunneling, reduce frequent node switching, and avoid multiple logins/logouts in a short time. If you’re on a company network, a security gateway may also return 403; in that case, verifying on a home network is the most straightforward.
429 rate limiting, interrupted output: treat “usage and concurrency” as the top suspect
If you encounter 429 during ChatGPT troubleshooting, it’s usually rate limiting caused by requests being too fast, concurrency being too high, or repeated submissions in a short period; stop for 1–3 minutes and try again—this is more effective than repeatedly clicking. If you have multiple tabs open, multiple devices, or you keep resending the same question, you may increase the chance of rate limiting; it’s recommended to keep only one chat window. If output is interrupted or gets stuck, first copy the content already generated, refresh the page, and then continue, and try to split long tasks into multiple rounds of conversation—this is also a practical ChatGPT troubleshooting habit.
Still unstable: do a final round of elimination along the “network path”
If you’ve completed all the above ChatGPT troubleshooting steps and it’s still unstable, suspect DNS, proxy split-tunneling, or firewall policies: temporarily switch back to the system default DNS, disable custom split-tunneling rules, or compare by switching to full global direct connection. Some routers’ parental controls/security filtering can also affect connectivity; reboot the router and turn off related filtering features, then verify again. If the issue occurs only on a specific device, first check whether the system time is correct and whether certificates/security software are blocking HTTPS—these details are often the final “real culprit.”
