If you’re searching “why is ChatGPT not working”, you usually want one of two answers:
Is it down for everyone? (OpenAI outage / incident)
Is it just me? (browser, network, account/session)
This guide gives you a fast troubleshooting checklist, then shows an optional automation: a Zo Agent that watches the official OpenAI status page and alerts you only when something changes.
Step 1: Check the official OpenAI status page (fastest signal)
OpenAI publishes incident updates here:
https://status.openai.com/
If ChatGPT is degraded or down, you’ll typically see an incident posted (or component degradation).
OpenAI’s own troubleshooting guidance starts with the same step: check status.openai.com, then refresh/clear cache, try private browsing, etc. ^1
Step 2: Quick “is it my browser?” triage
These take ~2 minutes and resolve a large fraction of “ChatGPT not loading / stuck” failures:
Hard refresh the page.
Try a private/incognito window.
If it works there, your issue is usually cookies, cache, or an extension.
Disable extensions (especially ad blockers, script blockers, privacy extensions), then reload.
Clear site data for ChatGPT (cookies + cached data), then sign in again.
If you’re seeing repeated “Something went wrong” / “Error in message stream” style errors, OpenAI’s help center recommends the same set of actions (refresh, check status, clear cache/cookies, private browsing). ^1
Step 3: Rule out network and VPN issues
If the status page is green but ChatGPT still fails:
Turn off VPN / proxy temporarily.
Switch networks (Wi‑Fi → hotspot) to test if your IP/network is flagged or unreliable.
If you’re on a restrictive corporate network, try from a personal connection.
A simple mental model:
Works on another network → your network/IP/policy is the culprit.
Works in incognito but not normal mode → cookies/extensions.
Fails everywhere, and status page shows incident → it’s OpenAI.
Step 4: When it’s an outage, have a fallback plan
If you rely on ChatGPT for work, outages are the moment you discover whether you have options.
Two practical fallbacks on Zo:
Use another hosted model provider (Claude, Gemini, etc.) for interactive work.
Run a local model for “good-enough” drafting / extraction / automation when APIs are flaky.
If you want a concrete local option, see:
https://www.zo.computer/tutorials/how-to-run-a-local-llm-on-your-server
Optional: Set up a Zo Agent to alert you when ChatGPT goes down (or recovers)
This pattern is useful because you don’t want to manually check the status page every time ChatGPT feels weird.
What this Agent does
On a schedule (e.g. every 5 minutes), it:
Fetches
https://status.openai.com/Extracts a small “status snapshot”
Compares it to the last snapshot saved in a file
Alerts you only when something changes
Create the Agent
Create a new Agent and use an instruction like this (edit the schedule and delivery to match your preference):
Fetch
https://status.openai.com/.Extract a short plain-text status summary:
overall banner status text (e.g. “We’re not aware of any issues…”)
any current incident title(s) and last update time
Read file
Status/openai-status-last.txtif it exists.If the summary is different from the previous run:
Write the new summary to file
Status/openai-status-last.txtSend me a short alert with:
“OpenAI status changed”
the previous summary (quoted)
the new summary (quoted)
If it’s the same, do nothing.
Tip: if you want screenshots (sometimes status pages are JS-heavy), use the browser-rendering tool instead of fast extraction. Zo tool reference:
https://docs.zocomputer.com/tools/read-webpage
https://docs.zocomputer.com/tools/view-webpage
For a broader “how Agents work” overview:
https://docs.zocomputer.com/agents
Create the state file folder (optional)
If you keep state in a dedicated folder, it’s easier to find later. Create a folder named Status/ in your workspace, and your Agent will store openai-status-last.txt there.
Summary
When ChatGPT isn’t working: