Is the AI service down, at its usage limit, or account-suspended? How to tell them apart
A direct answer to the question users ask when an AI tool stops working, distinguishing provider outage, plan usage limit, and account suspension by their evidence patterns, with the correct first action for each.
The question: 'the AI stopped working'—but this single observation has at least three distinct causes, each with different evidence, recovery steps, and urgency. Treating them all the same leads to wasted payments (buying another plan during an outage), duplicate tickets, or incorrect appeals. The short answer is to classify by who is affected and what the evidence says.
Provider or regional outage: many users fail simultaneously, and the official status page shows an incident. The evidence is the timing, region, endpoint, and error code. The correct action is to save the evidence and wait for updates—do not buy another plan or submit a refund request during an active outage, because the issue is transient and will resolve without payment changes.
Plan usage limit: only your account reports a message, token, credit, or rate-limit message while other functions work. The evidence is the specific limit type, the reset cycle, the model multiplier, and the usage dashboard. The correct action is to check when the limit resets, whether a higher tier or API plan is appropriate, and whether the usage is expected or indicates a leaked key. Account suspension: you cannot sign in, received a suspension email, or see a policy notice. The evidence is the email sender domain and the stated reason. The correct action is the official appeal flow, not repeated payment attempts.
When the cause is unclear, follow the sequence: official status page, then account usage dashboard, then plan and billing settings, then official email, then a support ticket. Describe a temporary 429, insufficient balance, or outage accurately—do not call every error 'banned' or 'suspended,' because the wrong label sends you down the wrong recovery path.