A ChatGPT internal server error usually means the request failed somewhere inside the service path, not that your prompt is permanently broken. Treat it as a server-side or connection-layer failure first, then rule out local causes such as stale site data, extensions, VPNs, proxies, secure DNS filters, and a stuck conversation. OpenAI’s own troubleshooting guidance recommends checking the status page, refreshing or starting a new chat, clearing cache and cookies, using an incognito window, disabling extensions, turning off VPN or secure DNS tools, and switching networks when needed.[1] This guide gives you the fastest order to test those fixes without wasting time.
What a ChatGPT internal server error means
A chatgpt internal server error is a broad failure message. In web terms, a 500 Internal Server Error means the server encountered an unexpected condition and could not fulfill the request.[3] In ChatGPT, the visible wording may vary. You may see “internal server error,” “something went wrong,” “there was an error generating a response,” a blank response, or a failed message after the spinner runs.
The key point is that the phrase does not identify one exact cause. The failure can happen in OpenAI’s service, in the model-serving path, in your browser session, in authentication, in a workspace feature, or in the network path between your device and ChatGPT. OpenAI’s help article describes ChatGPT errors as sometimes caused by network configuration, browser extensions, or transient server-side problems.[1]
That is why the right fix is not “clear everything” first. The right fix is to separate a broad outage from a local session problem. If many users are affected, local changes will not solve it. If only your browser profile is affected, waiting for OpenAI will not solve it. The steps below move from least disruptive to most diagnostic.
If you are seeing a plain HTTP-style page with 500, use the more specific ChatGPT error 500 internal server guide after this checklist. If the page says the service is unavailable rather than internal server error, compare it with chatgpt error 503 service unavailable.
Quick triage before you change settings
Start with a short triage pass. It tells you whether to troubleshoot your device or wait for recovery.
- Check OpenAI status. Open the OpenAI status page and look for active ChatGPT incidents. The status page also notes that availability metrics are reported at an aggregate level across tiers, models, and error types, so your experience can differ by plan, model, or feature.[2]
- Ask someone else or test another network. If the same prompt fails on your home Wi-Fi but works on cellular, suspect the network path, VPN, DNS, or filtering layer.
- Try a private window. Incognito or private browsing gives you a clean browser session. If ChatGPT works there, your regular browser profile is the suspect.
- Start a new chat. A single conversation can get stuck even when the service is otherwise fine.
- Check whether the error is feature-specific. File upload, image generation, voice mode, Custom GPTs, and Projects can fail while ordinary text chat still works.
This order matters. It avoids destructive fixes before you know what failed. For example, you do not need to wipe your browser history if the status page shows an active ChatGPT incident. You also should not assume a platform outage if ChatGPT works in a private window with extensions disabled.
If the symptom is slow responses rather than a hard error, read why ChatGPT is so slow. If the response starts and then breaks mid-stream, compare it with chatgpt error in message stream.
Step-by-step fix
Use this order. Stop when ChatGPT works again. Each step isolates a different layer of the problem.
Refresh, then retry once
Reload ChatGPT and send the prompt again. If the message was long, send a shorter version first. A one-time failure can happen when a request hits a bad backend path, times out, or loses the response stream. OpenAI’s guidance includes refreshing the page, starting a new chat, and using regenerate for failed responses.[1]
Check the status page before changing your setup
If OpenAI lists an active ChatGPT incident, do not keep changing your browser, password, or DNS. Wait, then retry. Status pages are not perfect for every individual account, but they are the best first signal for widespread elevated errors.
Start a new conversation
Copy your last prompt into a new chat. This tests whether the failure is attached to one thread. Long conversations, file-heavy threads, and tool-heavy sessions can fail while a clean chat still works. If a new chat works, keep using the new thread and archive the broken one later.
Open ChatGPT in a private window
Private browsing helps you test cookies, cached site data, and extensions without deleting anything. If ChatGPT works in private mode, return to your normal browser and disable extensions that block scripts, modify pages, filter traffic, or inject ChatGPT enhancements. OpenAI specifically calls out privacy and security tools, browser extensions, VPNs, proxies, and secure DNS tools as possible sources of ChatGPT errors.[1]
Clear ChatGPT site data, not your whole browser
If private mode works but your normal browser still fails, clear site data for ChatGPT-related domains rather than wiping every saved login in your browser. Then sign in again and retry. This is especially useful when the page loads but messages fail, history will not open, or the app shows repeated generic errors.
Turn off VPN, proxy, secure DNS, and filtering tools
VPNs and filtering layers can break the connection path even when ordinary websites work. Turn them off temporarily. Then reload ChatGPT and try again. If the error disappears, switch VPN locations, allowlist ChatGPT in the filtering tool, or use a different network. For a deeper pass, use our ChatGPT not working with VPN troubleshooting guide.
Switch device, browser, and network
Try a different browser, then a different device, then a different network. This tells you where the fault lives. If the same account fails everywhere, the issue may be account, plan, model, or service related. If it fails only on one browser, keep the fix local. If it fails only on one network, involve your network administrator or ISP.
How this error differs from nearby ChatGPT errors
ChatGPT error labels overlap. Use the wording and timing to choose the right fix. The table below gives the practical difference.
| Error wording | Likely meaning | Best first action | Related guide |
|---|---|---|---|
Internal server error or 500 | The server hit an unexpected condition and could not complete the request.[3] | Retry once, check status, then isolate browser and network. | chatgpt error 500 internal server |
| Something went wrong | A general ChatGPT failure that can be temporary server trouble or a local setup issue.[1] | Refresh, new chat, status page, cache, incognito, extensions, VPN. | chatgpt something went wrong error |
Bad gateway or 502 | A gateway or upstream service path failed. OpenAI has previously reported large portions of requests failing with 502 and 503 during a major service event.[8] | Check status and wait if the failure is widespread. | ChatGPT Error 502 Bad Gateway |
Service unavailable or 503 | The service path is unavailable or overloaded. Treat it as a capacity or service-side condition first.[8] | Wait, then retry. Avoid repeated refresh loops. | chatgpt error 503 service unavailable |
Too many requests or 429 | OpenAI says rate limit errors are caused by hitting request or token limits, and recommends retrying with exponential backoff for API use.[7] | Slow down requests and wait before retrying. | ChatGPT Error 429 Too Many Requests |
| Network error | Your device could not keep a reliable connection to OpenAI’s servers.[1] | Disable VPN or proxy, try another browser, and switch networks. | chatgpt network error |
If the error appears only when you upload a document, do not treat it as a general internal server error. Use the file-specific steps in chatgpt file upload not working. If it appears after a login redirect or authentication screen, start with chatgpt login failed.
When waiting is the correct fix
Waiting is correct when the error matches an active incident, affects multiple devices and networks, or appears suddenly across unrelated features. OpenAI has documented incidents where elevated errors affected ChatGPT conversations, login, account creation, Custom GPTs, Projects, and other service paths.[4][5]
Do not repeatedly hammer the send button during a service-side problem. Repeated retries can add load and make the failure feel worse from your side. In one OpenAI write-up, increased retry traffic during a disruption amplified downstream load and slowed recovery in one region.[5]
Use a measured retry pattern instead. Save your prompt outside ChatGPT. Wait until the status page changes, then retry in a new chat. If the work is urgent, switch to a lower-risk path: shorter prompts, fewer attachments, a different model if available, or another device on a separate network. If you track incidents for a team, bookmark chatgpt outages 2026 and the broader ChatGPT Outage History.
Special cases in the app, files, GPTs, and long chats
Mobile app errors
If the ChatGPT app shows an internal server error, test the web version before reinstalling. If the web version works, update the app, force close it, sign out and back in, and test on cellular instead of Wi-Fi. If both app and web fail on the same account, suspect service, account, or model availability rather than the app itself. For app-only symptoms, use chatgpt app not working.
File upload and data analysis errors
File workflows add upload storage, file processing, tool sessions, and generated downloads. A failure in any of those layers can look like a generic internal error. Try a smaller file, a fresh chat, and a different browser. If the issue appears after a tool session has been idle, compare it with code interpreter session expired.
Custom GPTs and Projects
Custom GPTs and Projects can have their own error patterns. OpenAI’s status history includes a resolved degraded-performance incident where users could see Internal Server Errors when editing or creating Custom GPTs and Projects.[4] If ordinary chat works but GPT editing fails, avoid changing your whole browser setup until you check whether that specific feature is degraded.
Very long conversations
Long chats can accumulate instructions, attachments, tool calls, and context that make failures harder to reproduce. Start a new chat with only the current task. If the new thread works, summarize the old thread and continue there. If memory or saved context seems involved, review chatgpt memory limit separately.
What to send OpenAI support if it keeps happening
Contact support when the error persists across browsers, devices, and networks while the status page does not show a matching incident. OpenAI asks users with persistent issues to collect a HAR file, browser console errors, timestamps, the model used, and the conversation URL or ID before contacting support.[1]
Send a concise report. Include the exact wording of the error, whether it happens in a new chat, whether it happens in private browsing, whether VPN or proxy is off, and whether another network works. Mention the feature: normal chat, file upload, image generation, voice, Custom GPT, Project, or login. If the problem is image-specific, use chatgpt image generator not working. If it is voice-specific, use chatgpt voice mode not working.
Do not send passwords, payment card numbers, private API keys, or unrelated browser history. A focused report is more useful than a long description. The goal is to give support enough evidence to separate account, browser, network, and service causes.
Frequently asked questions
Is a ChatGPT internal server error my fault?
Not usually. It often means a request failed somewhere in the server or service path. Your local setup can still trigger or expose the error, so you should test private browsing, extensions, VPN, and another network before assuming an outage.
Should I clear all browser history to fix it?
No. Start with a private window. If that works, clear ChatGPT site data or cookies first instead of deleting all browser history. This preserves unrelated logins and still tests the most likely browser-session problem.
Why does ChatGPT work on my phone but not my computer?
That usually points to a browser, extension, device, or network difference. Your phone may be using cellular data while your computer uses Wi-Fi, VPN, proxy, or a filtering DNS service. Test the same account in a private window and on a different network.
Why does only one conversation fail?
The conversation may be stuck because of its length, attachments, tool calls, or a failed response state. Copy the current prompt into a new chat. If the new chat works, continue there and avoid spending time on global troubleshooting.
Can a VPN cause an internal server error?
Yes. VPNs, proxies, secure DNS tools, and traffic filters can interfere with ChatGPT connections. Turn them off temporarily and retry. If that fixes the issue, switch VPN regions or allowlist ChatGPT-related traffic.
When should I stop troubleshooting and wait?
Stop when the same error appears across devices, browsers, and networks, or when OpenAI lists a relevant incident. At that point, local fixes are unlikely to help. Save your prompt and retry after the incident changes or resolves.