ChatGPT Keep Getting Error in Message Stream? How to Fix It

If ChatGPT keeps showing “Error in message stream”, the reply was cut off while it was being streamed to you, and the cause is usually on OpenAI’s side, in that particular conversation, or on your connection. The quickest fixes, in order: click Retry once, start a new chat if the same conversation keeps failing, check OpenAI’s status page, then turn off your VPN and browser extensions and try again.

Below is what the error actually means, how to work out which of those causes you’re dealing with, and what to do if none of the quick fixes stick. This reflects what’s known as of October 2026; OpenAI changes ChatGPT often, so details may shift.

What “Error in message stream” actually means

ChatGPT doesn’t send its answer all at once. It streams the text to your browser or app piece by piece over a connection that stays open while the model works. “Error in message stream” appears when that stream breaks before the answer is complete, which is why you often see a half-written reply with a Retry button underneath.

The message tells you where the failure happened, not why. According to a 2026 troubleshooting guide from AI Toolbox, OpenAI support has described it as a generic error with several possible causes. That’s why no single fix works for everyone.

Older versions of ChatGPT showed a similar message, “error in body stream”. Android Authority’s guide to that error pointed to the same broad set of causes: an unstable network, problems on OpenAI’s servers, or a prompt the system couldn’t handle.

The most common causes

Based on OpenAI’s help article and reports on the OpenAI Developer Community, the causes fall into four groups:

Server-side problems

When OpenAI has an incident, streams can break for large numbers of users at once. AI Toolbox notes that at least one past status incident was named after this exact message and was resolved by OpenAI rolling back a change. Nothing you do locally fixes that kind of problem; you have to wait it out.

A conversation that has gone bad

This is the pattern that matches “keep getting” the error. In one Developer Community thread, a user describes a regular chat with 30+ requests where continuing or branching the old conversation failed immediately, while a brand-new chat worked fine.

Length alone doesn’t seem to be the explanation, though. A separate thread about ChatGPT Projects reports repeated failures in Project chats and branches with a reasoning model, sometimes alongside “Something went wrong” and “Thinking failed” messages. OpenAI hasn’t published an official cause for these reports.

Your connection, VPN or browser

OpenAI’s help article on ChatGPT error messages says network errors mean your device can’t keep a stable connection to OpenAI’s servers. It recommends turning off VPNs, proxies and security filters, and trying another browser or network. Because a streamed reply needs the connection to stay open for the whole answer, anything that interrupts or inspects long-running connections can break it.

Attachments

A community report describes the error appearing only when image files were attached, while PDF and JSON files worked. No fix was posted in that thread, so treat this as a pattern to test rather than a confirmed bug.

How to fix it, step by step

Work through these in order. Most people are sorted by step 3.

1. Click Retry once or twice. A brief hiccup often clears on its own. Don’t spam it; if it fails repeatedly in the same chat, move on.
2. Wait 30-60 seconds if the reply seems frozen. OpenAI’s help article suggests waiting that long when a response hangs, then using stop and regenerate.
3. Start a new chat. If your current conversation fails every time, open a fresh chat and paste in the context you need (a short summary of the earlier conversation usually works better than pasting the whole thing). This is the most consistently reported workaround.
4. Check status.openai.com. If ChatGPT is listed with an incident or degraded performance, the problem isn’t on your end. Wait and try again later.
5. Turn off your VPN, proxy or security filter temporarily, as OpenAI’s help article recommends, and try again. Corporate networks and antivirus “web shield” features can have the same effect.
6. Test in a private/incognito window. This runs ChatGPT without most extensions and with fresh site data. If it works there, an extension (ad blockers and privacy tools are common suspects) or stale cache is likely the cause.
7. Clear ChatGPT’s cache and cookies in your browser, then sign back in. OpenAI’s article lists this among its fixes for general errors.
8. Try a different browser, device or network. Switching from Wi-Fi to mobile data, or from the browser to the official ChatGPT app, quickly shows whether the problem follows your account or your setup.
9. Remove or change the attachment. If the error started after you uploaded an image or file, send the message without it, or try a different file format, to see if the upload is the trigger.
10. Simplify the request. If you’re asking for a very long answer or a heavy reasoning task, break it into smaller parts. Guides covering the older “body stream” error also suggest shortening or rephrasing the prompt.
11. Sign out and back in. OpenAI’s help article lists this as a step if starting a new chat doesn’t help.

If you’re a developer using the Apps SDK or API

The same message also shows up in developer contexts. There’s an OpenAI Developer Community thread about it appearing in the ChatGPT Apps SDK development app, and several API threads about image attachments failing. In one of those, an OpenAI staff member asked affected developers to open a support request and include a HAR file (a browser log of network requests). If you’re building on OpenAI’s platform, that’s the route to take, since it gives support something concrete to investigate.

When to contact OpenAI support

Contact support through the official OpenAI Help Center if:

• the error happens in every new chat, on more than one device and network, and
• status.openai.com shows no incident, and
• you’ve already tried a private window and turning off your VPN.

At that point the problem is probably tied to your account or a specific feature rather than your setup. Describe which model you were using, whether it happens in Projects or regular chats, whether attachments were involved, and roughly when it started. Never share your password with anyone claiming to offer help; OpenAI support doesn’t need it.

The bottom line

“Error in message stream” is a catch-all for a reply that stopped arriving partway through. If it keeps happening in one conversation, the fix is almost always a new chat. If it’s happening everywhere, check OpenAI’s status page first, then rule out your VPN, extensions and network. OpenAI hasn’t published a definitive list of causes, and community reports suggest more than one thing can trigger it, so working through the steps above in order is the most reliable approach.