Why an AI service can be “up” but broken for you

“All systems operational” is a statement about the provider’s platform, not about your session. There are several ways to be left out of a working service.

The status page is slower than the problem

Providers often confirm an incident only after their own engineers have investigated. In the first minutes of a real outage, a status page can still show green. A status checker like ours narrows that gap a little, because our own request does not wait for a human to update a page, but it cannot close it entirely.

Only part of the service is affected

A provider can report a problem with voice mode, file uploads or one model while the main chat works. Whether that matters depends on what you were trying to do. On each service page we list the components the provider says are affected, not only the headline.

It is specific to your account or plan

Usage limits, a lapsed payment, a flagged account or a feature that has not reached your plan can all produce errors that nobody else sees. These never appear on a status page.

It is specific to your region or network

Traffic is routed through different data centres in different places. A fault in one region, or a block applied by your employer, school or internet provider, can leave you cut off while the rest of the world carries on.

Your browser is in the way

Extensions, stale cookies, a VPN, or a bot-protection check that keeps failing can make a working site look dead. A private window on a different network settles it quickly.

Demand is the problem

At busy times a service can be technically up but so loaded that requests time out or queue. Providers do not always class that as an outage, although it feels like one.

To test which case you are in, follow the steps in what to do when an AI chatbot stops working.