Last updated: May 9, 2026 · Verified against ChatGPT-5 and GPT-4o
ChatGPT problems come in several distinct flavors. Maybe the page loads but you see "Something went wrong" every time you send a message. Maybe responses start generating and then abruptly cut off mid-sentence. Maybe the chat window loads but the send button does nothing. Or maybe ChatGPT just refuses to load at all, sitting on a spinning indicator indefinitely. Each of these has a different cause and a different fix.
Before diving into fixes, it helps to know that ChatGPT's problems divide into two camps: problems with OpenAI's servers (which you can't fix — you can only wait or work around them), and problems with your browser, app, or account (which you absolutely can fix). The quick diagnosis below helps you figure out which category you're in within 30 seconds.
ChatGPT not working for everyone → Server/outage issue. Check status.openai.com first. Nothing you can do but wait.
ChatGPT works in Incognito but not your normal browser → Browser extension or cookie issue. See Fix 2.
ChatGPT works on mobile app but not in browser (or vice versa) → Browser-specific issue or app-specific bug. See Fix 3 or Fix 4.
"Something went wrong" appears after login → Session/cookie issue. See Fix 2.
Responses get cut off → Connection timeout. See Fix 5.
"You've reached your limit" message → Free tier rate limit hit. See Fix 7.
Can't log in at all → Account issue. See Fix 6.
Why this works: OpenAI's servers experience outages, slowdowns, and maintenance windows. Spending 20 minutes troubleshooting your browser when ChatGPT is down site-wide is wasted effort. The status page takes 5 seconds to check and immediately tells you if the problem is on their end.
If there's an active incident: your fix is to wait. OpenAI typically resolves issues within 30 minutes to a few hours. If the status page shows everything operational, your problem is local — continue to the next fixes.
Why this works: ChatGPT uses browser cookies to maintain your login session and store UI state. If these cookies become corrupted — from a browser update, an OpenAI session change, or just normal browser database corruption — you might be stuck in a broken authenticated state: ChatGPT thinks you're logged in but with an invalid session, causing "Something went wrong" on every request.
Faster method: Open chat.openai.com, press F12 to open Developer Tools → Application tab → Cookies → right-click openai.com → Clear. Then refresh and log in.
Also try opening ChatGPT in an Incognito or Private browser window — if it works there, a cookie or extension issue is confirmed.
Why this works: Some browser extensions interfere with ChatGPT's JavaScript, particularly extensions that modify page content or block network requests. Ad blockers like uBlock Origin, privacy extensions like Privacy Badger, and some password managers have been known to break ChatGPT's interface or prevent requests from reaching OpenAI's API.
Why this works: ChatGPT uses modern web APIs including WebSockets for real-time streaming. Outdated browsers may not support these correctly. Specific browser bugs can also affect ChatGPT — Chrome sometimes handles streaming responses differently than Firefox or Edge.
If you're using Chrome, try Firefox or Microsoft Edge. If you're on Safari, try Chrome. A working alternative browser immediately confirms your primary browser is the issue.
To update Chrome: click the three-dot menu → Help → About Google Chrome — it'll update automatically and prompt you to relaunch. For Safari on iPhone: go to Settings → General → Software Update to update iOS, which updates Safari.
Why this works: ChatGPT streams responses in real-time using a persistent HTTP connection. Any interruption in your internet connection — even a brief dropout of less than a second — can break the stream and end the response abruptly. This is different from the response being "censored" or hitting a content limit; those look different.
First, try to continue the response by typing "please continue" or "finish your previous response." If ChatGPT can see the partial response in context, it will complete it.
To prevent future cutoffs:
Why this works: ChatGPT login issues are often caused by signing in the wrong way — using email/password when you originally signed up with Google, or vice versa.
Why this works: Free ChatGPT users have usage limits. When you hit them, you see a message like "You've sent too many messages. Please wait until [time] to send a message." This isn't a bug — it's an intentional limit on the free tier.
Free tier limits: GPT-4o has a usage cap that refreshes every few hours. When reached, ChatGPT falls back to GPT-4o mini or GPT-3.5 automatically — you'll see a notice at the top of the chat window when this happens.
Options when you hit the limit:
Why this works: The ChatGPT mobile app (iOS and Android) and the web version don't always experience issues simultaneously. The app communicates directly with OpenAI's API, bypassing browser caching, cookies, and extensions entirely. When the web version is having problems, the app often works fine.
Download on iPhone: Open the App Store and search "ChatGPT" — install the official app by OpenAI, Inc. (the developer name, not just "OpenAI").
Download on Android: Open Google Play Store and search "ChatGPT" — install the app from OpenAI. Be careful of copycat apps; the official publisher is "OpenAI."
Once installed, log in with the same account you use on the web. Your conversation history syncs automatically between the app and web versions.
If ChatGPT is showing a specific error rather than just "not working," the exact wording tells you what's wrong. Here's what each common 2026 error means and the fix for it.
What it means: The connection between your browser and OpenAI's servers dropped while ChatGPT was sending you a response. The stream broke mid-flight, leaving the message incomplete or empty.
Fix: Click the regenerate button (circular arrow under the broken message) or type "please continue from where you left off." If it happens repeatedly, your connection is unstable — switch from Wi-Fi to wired, or try the mobile app, which handles stream drops more gracefully than the browser.
What it means: Same root cause as the message stream error — the stream was cut, but ChatGPT is now waiting to see if it can resume. Usually it can't.
Fix: Wait 30 seconds. If the message doesn't complete, regenerate. If you see this repeatedly within one session, OpenAI is having capacity issues — check status.openai.com.
What it means: ChatGPT can't fetch your chat history from OpenAI's servers. This is almost always a session/auth problem, not a server outage.
Fix: Refresh the page once. If that fails, log out completely (avatar → Log out) and log back in. If still broken, clear cookies for openai.com — your session token is corrupted.
What it means: Generic catch-all error from ChatGPT's frontend. The actual cause is one of: expired session, blocked request from a browser extension, OpenAI API hiccup, or rate limit hit.
Fix: Try in an Incognito window first (this rules out extensions in 5 seconds). If Incognito works, an extension is your problem (see Fix 3). If Incognito also fails, clear cookies and log back in. If still broken, OpenAI is likely having a server-side issue.
What it means: Either you've hit your GPT-5 message limit (free and Plus users have separate caps), or GPT-5 is experiencing higher load than other models. ChatGPT silently falls back to GPT-4o or GPT-4o mini in some cases.
Fix: Check the model selector at the top of the chat — if it shows GPT-4o instead of GPT-5, you've hit your cap and the fallback kicked in. Wait for the reset window (shown in the notification banner) or upgrade to a higher tier. If GPT-5 is selected but slow, switch to GPT-4o temporarily — it's often faster during peak hours.
What it means: The voice/audio playback failed — either the audio file didn't generate on OpenAI's side, or your browser blocked autoplay.
Fix: Click the play button manually (browsers block automatic audio in many cases). If that doesn't work, refresh the page — voice generation can fail intermittently when OpenAI's TTS service is overloaded. Check site permissions: in Chrome, click the lock icon next to chatgpt.com → Site settings → ensure Sound is allowed.
What it means: The DALL-E backend rejected your prompt or experienced an error. Most often it's the safety filter overblocking, but it can also be a server hiccup.
Fix: Rephrase the prompt — even reworded versions of the same concept often pass the filter on the second try. Avoid named real people, brand logos, and anything that could be interpreted as violent or sexual. For more detail see our DALL-E image generation failed guide.
What it means: The conversation ID in your URL no longer exists on OpenAI's servers — usually because it was archived, you're looking at a shared link that was deleted, or there's a brief sync issue.
Fix: Go back to chat.openai.com (the root) — your active conversations should still be in the sidebar. If a specific conversation has actually disappeared, check the Archived section under Settings → Personalization. OpenAI doesn't permanently delete chats unless you explicitly do so.
What it means: Rate limit on requests-per-minute, separate from the daily message cap. Triggered if you hit regenerate repeatedly or send long messages back-to-back.
Fix: Wait 60 seconds and try again. If it persists, OpenAI's anti-abuse system has flagged your IP — common on shared networks (offices, schools, public Wi-Fi) or if you're on a VPN. Try disconnecting the VPN or switching networks.
Q: Why does ChatGPT say "Something went wrong" every time I send a message?
A: This error typically means one of three things: OpenAI's servers are under high load (check status.openai.com), your browser session has expired or become corrupted (clear cookies for openai.com and log back in), or a browser extension is interfering with ChatGPT's JavaScript. Try loading ChatGPT in an Incognito window with extensions disabled — if it works there, an extension is the culprit.
Q: Why are ChatGPT responses getting cut off mid-sentence?
A: Responses cut off mid-generation usually mean a connection timeout due to an unstable internet connection, or the conversation hit a very long context. Type "please continue" to resume. If cutoffs happen frequently, check your connection stability — ChatGPT streams responses in real-time, and even a brief dropout stops the stream. Switching to a wired connection or trying the mobile app often resolves this.
Q: ChatGPT is extremely slow to respond — what's wrong?
A: Slow ChatGPT responses are almost always a server load issue on OpenAI's side, not your device or connection. This is especially common during peak usage hours in the afternoon and evening in North America. ChatGPT Plus subscribers get priority access during high-traffic periods, meaning significantly faster responses. If you use ChatGPT for work, the Plus tier is worth considering.
Q: Why can't I log into ChatGPT even though my password is correct?
A: The most common cause is using the wrong login method. If you originally signed up with Google or Apple, you must use "Continue with Google" or "Continue with Apple" — there's no email/password for those accounts. If you signed up with email, use the email/password flow. If that still fails, clear your browser cookies for openai.com and try again, or use the Forgot Password link.
Q: Why does ChatGPT work in the app but not in my browser?
A: Browser extensions, corrupted cookies, or an outdated browser version are the most common causes of browser-specific ChatGPT failures. Try opening ChatGPT in an Incognito window (extensions are disabled by default). If it works in Incognito, an extension is interfering. If it works in a different browser entirely, your primary browser needs its cookies cleared or an update.
If ChatGPT still isn't working after trying all these fixes, and OpenAI's status page shows everything is operational, your account may have a specific issue. Contact OpenAI's support team at help.openai.com. Plus and Team plan subscribers have priority support queues. You can also check the OpenAI Community Forum for reports of issues affecting specific accounts or regions.