ChatGPT, Claude or another AI tool not working? What to check

AI chat tools are some of the busiest services on the internet. When one stops answering, it is usually capacity or a service fault, not your device, but a few local causes are worth ruling out first.

What the symptoms usually mean

  • Endless spinner or “something went wrong”: the service accepted your request but could not complete it. Typical of overload or a partial outage.
  • “At capacity” or a queue message: too many people at once. It clears as demand drops.
  • Replies stop halfway: the connection dropped while the answer was being streamed. Your network, a VPN, or a server restart can cause it.
  • Error 429: you or your network sent too many requests. See error 429.
  • Error 5xx: the service itself failed. Read the server error guide.
  • Login loops: often a stale cookie or a problem at the sign-in provider rather than the chat tool itself.

Why AI services are fragile under load

Each answer needs a lot of computing power, and that cannot be cached the way a web page can. A sudden surge, for example after a new model launch or a news story, can use up capacity. The website, the sign-in system and the model that writes the answers are separate parts, and any one of them can fail while the others work. That is why a site can load fine while the chat itself errors.

Step by step

  1. Run the tool’s address through the checker. If it fails from many places, it is the service.
  2. Check the provider’s own status page. These often name which part is affected, such as the API, logins or one model.
  3. If the service is up, refresh once, then try a private window to rule out cookies and extensions.
  4. Turn off a VPN or ad blocker for a moment. Some AI sites treat VPN traffic as risky.
  5. Try another network, such as mobile data.
  6. Start a new chat. A very long conversation can hit limits that a new one does not.

A worked example

You send a prompt and see “network error” after ten seconds. The checker shows the website up but the status page lists “elevated errors” for the chat feature. That is a partial outage. Retrying repeatedly will not help and can trigger rate limits, so you wait and check again in 15 minutes. If the status page showed nothing, but the problem only happened on Wi-Fi, a router restart or a different DNS server would be the next test.

While you wait

  • Copy any long prompt you wrote somewhere safe so you do not lose it.
  • Avoid resubmitting the same request many times.
  • If you rely on the tool for work, keep an alternative ready and use the watchlist to get an email when it recovers.

If you build on an AI API

Add retries with increasing delays, set sensible timeouts, handle 429 and 5xx responses separately, and avoid sending bursts of requests. Monitoring your own endpoint, plus the provider’s, tells you quickly whether a failure is yours or theirs. See how to monitor your own website.

Check a website now

Is it down for everyone, or just you? Paste an address and we test it from fifteen places.

Open the checker

Related guides

Last updated 2 November 2026. Written by the siteupdown.com team. Spotted a mistake?

← All help guides