You’re mid-prompt, ChatGPT starts typing a reply, then it stops dead and throws up a red “Error in message stream” message. Your conversation halts, the response is half-finished, and you have no idea whether the problem is on your end or OpenAI’s.
Here’s the good news: this error is almost always fixable in under 60 seconds. The trick is trying the fixes in the right order — starting with the ones that work 80% of the time before touching anything technical.
This guide walks through 9 proven fixes, ordered from fastest to most involved.
What “Error in message stream” actually means
ChatGPT doesn’t send its replies in one chunk — it streams them word by word from OpenAI’s servers to your browser or app in real time. That’s why you see the text appear token by token instead of all at once.
The “Error in message stream” message means that stream got interrupted before the reply finished. Something — your network, OpenAI’s server, your browser, an extension, a VPN — broke the live connection between your screen and OpenAI.
Three important things to know before you start troubleshooting:
- The error is usually temporary. Most cases clear up within a minute or two just by retrying.
- It can be on either side. About 40% of the time it’s OpenAI’s servers; 60% of the time it’s something on your end.
- Partial replies still appear. If ChatGPT started typing before the error, that text is usually still there. Don’t lose it — you can copy it out before trying a fix.
The 9 fixes (try them in this order)
1. Click “Regenerate” (fixes ~50% of cases)
The simplest and most effective fix. Right below the cut-off response, ChatGPT shows a Regenerate button. Click it.
This re-sends the same prompt without starting a new conversation. For transient network blips or server hiccups — which is what most stream errors actually are — a single retry produces a successful response. If the first regenerate fails, try it once more. Don’t hammer it 10 times, though; if two retries don’t work, the problem is elsewhere.
2. Check OpenAI’s status page
Before you troubleshoot your own setup, confirm it’s not on OpenAI’s end.
Go to status.openai.com and look for any active incidents. If you see “degraded performance,” “partial outage,” or a mention of message stream errors, the fix is to wait. OpenAI typically resolves these within 10–60 minutes, and no amount of clearing your cache will help while their servers are struggling.
This is worth a 10-second check before you start making changes to your browser.
3. Start a new chat
Long conversations are one of the biggest triggers for stream errors. As a chat grows, ChatGPT has to load and re-send more context with every reply, which makes the stream more likely to fail at higher token counts.
Open a new chat and paste the same prompt in. If the new chat works, your old conversation had gotten too long — that’s your signal to break it into smaller chats going forward.
A rough rule of thumb: conversations above 30–50 messages start hitting these stream issues more often, especially on GPT-4o or GPT-5 with longer replies.
4. Refresh the page or restart the app
On web: press Ctrl + R (Windows) or Cmd + R (Mac) to reload ChatGPT.
On mobile: force-close the app (swipe it away from the app switcher) and reopen it.
This clears any stuck session state, re-authenticates you, and re-opens the WebSocket connection ChatGPT uses to stream responses. It takes 5 seconds and fixes a surprising number of cases.
5. Check your internet connection
Streaming requires a stable, continuous connection. If your Wi-Fi is dropping packets or your signal is weak, the stream will break even though basic web browsing still works.
Quick tests:
- Switch networks. Try the same prompt on your phone’s mobile data (turn off Wi-Fi). If it works there, your Wi-Fi is the problem.
- Restart your router. If everyone in the house is seeing slow internet, a router reboot fixes it in 2 minutes.
- Check for VPN interference. Turn your VPN off and retry. VPNs — especially free ones — are a notorious cause of broken streams because they drop idle connections and introduce latency.
- Run a quick speed test. Anything below 5 Mbps or above 200ms ping is going to strain the stream.
6. Disable browser extensions (especially ad blockers)
This is the single most common user-side cause of stream errors, and most troubleshooting guides bury it at the bottom.
Ad blockers, privacy tools, script blockers (uBlock Origin, AdBlock Plus, Privacy Badger, Ghostery), and some security extensions intercept or block parts of the WebSocket connection ChatGPT uses to stream. They weren’t designed to interfere — ChatGPT’s traffic just sometimes trips their filters.
The fastest way to test:
- Open ChatGPT in a private/incognito window (extensions are disabled by default there).
- Log in and send a prompt.
- If it works, extensions are the problem.
To fix permanently: go to your ad blocker’s settings and whitelist chatgpt.com and chat.openai.com. That’s usually a one-click option. Don’t disable the extension entirely — just allow ChatGPT.
7. Clear your browser cache and cookies for ChatGPT
Stale cached files or corrupted session cookies can silently break authentication mid-stream. Clearing them forces a clean reconnect.
Chrome / Edge / Brave:
- Press Ctrl + Shift + Delete (Windows) / Cmd + Shift + Delete (Mac).
- Select “Cookies and other site data” + “Cached images and files.”
- Set time range to “Last 24 hours” (faster than “All time”).
- Click “Clear data.”
- Log back into ChatGPT.
Firefox: Settings → Privacy & Security → Cookies and Site Data → Clear Data.
Safari: Settings → Privacy → Manage Website Data → search “openai” → Remove.
You’ll have to sign in again after clearing, but it resolves a meaningful share of persistent errors.
8. Switch browsers (the 10-second diagnostic)
If clearing the cache didn’t help, switch to a completely different browser — Edge if you were on Chrome, Firefox if you were on Safari.
If ChatGPT works in the other browser, the original browser has a profile-level issue: a corrupted cookie store, a broken extension, or a stuck update. Either clear it more aggressively or just keep ChatGPT in the second browser.
If ChatGPT also fails in the second browser, the issue isn’t browser-specific — move to the next fix.
9. Split up long prompts or large file uploads
The error fires disproportionately when you’re:
- Pasting a 5,000+ word prompt
- Uploading a large PDF, DOCX, or image file
- Asking for a very long response (ChatGPT’s output limit is fuzzy but reliable errors start appearing past ~3,000 words of continuous output)
ChatGPT’s streaming infrastructure has limits. Long inputs and outputs push against them, and when they break, you get exactly this error.
Fixes:
- Break large prompts into chunks. Send the first half, ask ChatGPT to wait, then send the second half.
- Split big files. If you’re uploading a 50-page PDF, extract the 5 pages you actually care about and upload just those.
- Ask for shorter outputs. “Give me this in 3 sections of ~500 words each, and wait for me to say ‘continue’ between sections.”
- Try a different model. If you’re on GPT-4o and hitting repeated stream errors on long tasks, switch to GPT-5 or a reasoning model — they handle long contexts more robustly.
Common variations of this error (same cause, different wording)
ChatGPT shows slightly different error text depending on where the stream broke. They’re all the same class of problem, and all the fixes above apply:
- “Error in message stream”
- “Error in body stream”
- “There was an error generating a response”
- “Something went wrong. Please try again.”
- “Network error” (with a retry button)
If you see any of these, walk the same 9-fix list.
How to prevent the error from happening in the first place
Once you’ve fixed the immediate issue, a few habits dramatically reduce how often it comes back:
- Keep chats under ~50 messages. Start a new conversation for each major topic. Pin important context with Custom Instructions instead of dragging it through endless long threads.
- Whitelist ChatGPT in your ad blocker permanently. Once you’ve done it, forget about it.
- Prefer Wi-Fi over cellular for long sessions — but not public Wi-Fi, which is often the worst offender.
- Keep your browser updated. Chrome, Edge, Firefox, and Safari all push updates that include WebSocket stability fixes.
- Avoid free VPNs entirely when using ChatGPT. They’re the single most common cause of persistent stream issues.
- Split big tasks into small ones. If you’ve got a 10,000-word document to process, feed it in pieces. Your success rate goes from 50% to 99%.
When to escalate (it’s probably not you)
If you’ve worked through all 9 fixes and the error keeps appearing:
- Check Downdetector at downdetector.com/status/openai for user-reported outages that haven’t made OpenAI’s official status page yet.
- Search X/Twitter for “ChatGPT down” — users spot outages there 20–30 minutes before official announcements.
- Contact OpenAI support through help.openai.com if the error persists for over an hour and no outage is reported. Include the time, your account email, and a screenshot of the error.
- If you’re on a corporate or school network and ChatGPT has been fine from home, your IT department is probably filtering or throttling the connection. That’s a conversation with IT, not something you can fix from the browser.
Frequently asked questions
Is “Error in message stream” my fault or ChatGPT’s fault? Roughly 40% of the time it’s OpenAI’s servers, 60% of the time it’s on your end (network, browser, extensions). Always check status.openai.com first — it’s free information that takes 10 seconds.
Will I lose my conversation when this error happens? No. The chat history is saved before the error appears. The specific reply that failed won’t be saved, but everything before it is intact. You can safely refresh or click regenerate.
Does the error happen more on free ChatGPT than Plus? Slightly. During server load spikes, free users are the first to experience stream errors and retries. Plus and Pro users get priority routing but aren’t immune.
Can using a VPN cause this error? Yes — this is one of the most common hidden causes. Free and low-quality VPNs drop idle connections, introduce high latency, and sometimes route ChatGPT through servers that OpenAI’s infrastructure flags. Turn your VPN off for ChatGPT, or upgrade to a premium provider with stable long-lived connections.
The error only happens when I upload files — why? File uploads push much more data through the stream than text prompts, and they hit processing limits that short prompts don’t. If you’re consistently erroring on uploads, try a smaller file, a different format (PDF usually works better than DOCX), or uploading one file at a time instead of several.
Does restarting my computer help? Almost never for this specific error. The problem is at the browser/network/server layer, not the OS. Skip the restart and work through the 9 fixes above.
Is there a ChatGPT error in message stream limit I’ll hit as a free user? There isn’t a specific error-message limit. But free users hit message caps, model downgrades, and server-deprioritization more often than paid users, all of which correlate with more stream errors.
The “Error in message stream” is almost always a minor hiccup that looks scarier than it is. Most of the time, Regenerate fixes it. When it doesn’t, walk the list in order — check the status page, start a new chat, check your connection, disable extensions — and you’ll clear the vast majority of cases in under two minutes.
If you’re running into other ChatGPT issues while you’re here, check our guides on uploading images to ChatGPT and changing your ChatGPT password — both cover adjacent problems that tend to show up in the same troubleshooting session.