AI Chatbot Errors Explained: Capacity, Rate Limits and Network
An AI chatbot error usually falls into one of three groups: your own usage limit, a temporary capacity squeeze, or a real outage. A usage limit means you used up your allowance and it resets at a stated time. A capacity message means demand is high across all users. An outage means something is broken and the provider’s status page says so.
Which one you have decides what helps. Waiting fixes the first two; only a provider fix solves the third; a network problem on your side looks like all three. Our status pages for ChatGPT, Claude, Gemini and Copilot show current availability.
What is a usage limit message?
A usage limit message means your account used its allowance for a period and must wait for the reset. It is a rule of your plan, not a fault. Anthropic’s Help Center article “Troubleshoot Claude error messages” lists wording such as “5-hour limit reached – resets [time]” and “You’ve reached your weekly limit. It resets [day] at [time].” Anthropic’s article says these limits appear as warnings first, then as a blocking error, and that plans with usage credits enabled can continue on pay-as-you-go pricing.
OpenAI says similar things for ChatGPT: when a limit is reached, access to the affected model pauses until the limit resets, and ChatGPT shows the reset timing in the product.
What does a capacity or “overloaded” message mean?
A capacity message means the service is working but cannot take your request at this moment because demand is high. Anthropic’s help article gives this wording: “Due to unexpected capacity constraints, Claude is unable to respond to your message. Please try again soon.” It describes the cause as high system-wide demand rather than a technical fault, advises waiting a few minutes, and notes that such a message does not appear on the status page.
That last point matters. A clear status page does not rule out a capacity message, and a capacity message does not mean an outage.
What do rate limit and “Too Many Requests” errors mean?
A rate limit error, shown to developers as HTTP 429, means too many requests or tokens were sent in a short window. OpenAI’s documentation says unsuccessful requests still count toward your per-minute limit, so sending again immediately makes it worse; it recommends exponential backoff, meaning waiting longer between each retry. Anthropic’s API documentation uses 429 for rate limits and a separate 529 overloaded_error for a temporarily overloaded API. OpenAI’s docs describe a 503 with the code server_is_overloaded in the same spirit. For what these numbers mean on ordinary websites, see HTTP error codes explained.
Chat app users rarely see these numbers. They matter if an app or tool built on an AI API stops working.
What does “network error” mean?
A network error means the connection between your device and the service broke or never formed. OpenAI’s help article says a network error or websocket connection error suggests your device cannot establish a reliable connection to its servers. Common local causes are a VPN, a proxy, secure DNS, a browser extension or an unstable Wi-Fi connection.
How do I tell the three apart?
| Message type | What it means | Appears on status page? | What to do |
|---|---|---|---|
| Usage limit | Your allowance is used up | No | Wait for the stated reset time, or change plan |
| Capacity / overloaded | Demand is high for everyone | Often not | Wait a few minutes, retry once |
| Rate limit (429) | Too many requests in a short time | No | Slow down; wait before retrying |
| Outage | Something is broken on the provider’s side | Usually yes | Watch incident updates |
| Network error | Your connection to the service failed | No | Switch network, turn off VPN, clear cache |
For a quick test, load the provider’s status page, then try the chatbot on mobile data. The full routine is in is it down or just me. Status labels are explained in how to read a status page.
What should I try, in order?
- Read the exact words of the message and note any reset time.
- Open the provider’s status page and our ChatGPT, Claude or Gemini page.
- Wait a few minutes and retry once. Repeated rapid retries can trigger rate limits, as OpenAI’s documentation notes for failed requests.
- Test a second network or device to rule out a network error.
- If a limit message names a reset time, plan your work around it.
What if an app built on an AI tool fails?
Many apps call an AI provider’s API behind the scenes. When they fail, the app shows its own error, and the cause may be a rate limit, an overloaded API or an outage at the provider. You cannot fix that from your side. Check the provider’s API status and the app’s own page, and wait.
Bottom line
Read the message first. A reset time means a limit, “try again soon” means capacity, and a clear connection failure means check your network. Only an incident on the provider’s status page confirms an outage.
FAQ
Does a capacity message mean the chatbot is down?
No. Anthropic’s help article describes it as the system working normally while it manages high demand. Wait a few minutes and try again.
Can I get around a usage limit?
Only in ways the provider offers, such as a higher plan or usage credits where available. Do not use unofficial tools or shared accounts; they risk your account and your data.
Why does the chatbot work on my phone but not my computer?
The service is up, so the cause is local: a VPN, extension, cached data or the network. Try a private window and another network.
Are these messages the same in Gemini and Copilot?
The wording differs by product, so read each provider’s help pages for exact text. The three-way split between limits, capacity and outages applies broadly, and the status page is the way to confirm an outage.