Your AI assistant freezes mid-sentence. A spinning cursor. A vague “something went wrong.” Before you refresh for the tenth time or assume your account is broken, take a breath. Most of the time the problem is fixable in under a minute, and when it isn’t, there’s a clear way to find out and a smart way to wait it out. This guide walks you through exactly what to do when an AI tool is down, from confirming the outage to getting back to work fast.
The 30-second version
- Confirm it’s actually down by checking a live status page like our AI status dashboard before troubleshooting.
- Rule out your side: your internet, browser, extensions, VPN, or account can all look like an outage.
- Read the official status page to see whether the provider has acknowledged a problem.
- Try the fast fixes: refresh, re-login, clear cache, switch network or browser.
- If it’s a real outage, subscribe to updates, switch to an alternative tool, and come back later.
Step 1: Confirm the AI tool is actually down
The first mistake people make is assuming the worst. An error message doesn’t always mean the service is offline. It can mean a single request failed, your session expired, or your network hiccupped. The fastest way to separate “the service is down for everyone” from “something is wrong on my end” is to check a third-party monitor that watches the service continuously.
Open our live status dashboard and find the tool you’re using. Each service has its own page with a real-time verdict, recent uptime, and a count of how many other people are reporting problems right now. If you see a wave of reports and a red or amber status, the issue is on the provider’s side and there’s nothing to fix on yours. If everything looks green and quiet, the problem is probably local, so move on to Step 2.
Jump straight to the service you need: Is ChatGPT down? · Is Claude down? · Is Gemini down? · Is DeepSeek down? · Is Grok down? · Is Copilot down? · Is Character.AI down?
Step 2: Rule out your own side first
A surprising share of “outages” are local problems wearing an outage costume. Before you wait around for a fix that will never come, because nothing is actually broken on the provider’s end, run through this short list. Each one takes seconds and resolves the majority of “it’s not working” moments.
- Your internet connection. Load any other website. If nothing loads, the AI tool isn’t the problem, your connection is. Toggle Wi-Fi off and on, or switch to mobile data.
- Browser cache and cookies. A stale session is a classic cause of stuck screens. Open a private window and sign in there; if it works, clear your normal browser’s cache for that site.
- Browser extensions. Ad blockers, script blockers, and privacy extensions sometimes break chat interfaces. Disable them for the site or test in incognito.
- VPN or region. Some AI services are restricted in certain countries or behave oddly through a VPN. Turn the VPN off and retry.
- Your account or login. A logged-out session looks like a dead app. Sign out fully and back in.
- The app version (mobile). Force-close the app, check for an update, and reopen it.
If you’ve worked through this and the tool still fails, yet the status dashboard shows green, the problem is almost certainly local. If the dashboard shows a real problem, skip ahead to Step 4.
Step 3: Check the official status page
Most AI providers publish an official status page where they post incidents and maintenance windows. It’s the authoritative source for whether the company has acknowledged a problem, but it has one weakness: it can lag. During the first minutes of an outage, the status page may still say “all systems operational” while users are already locked out. That gap is exactly why we run an independent reachability check alongside the official feed; you can read how that works on our methodology page.
When you read a status page, look for three things: an active incident (a red or yellow banner), the affected components (is it the chat interface, the API, logins, or image generation?), and the timestamp of the last update. A recent update means the team is actively working on it; a stale page during a clear outage usually means the incident hasn’t been posted yet.
Step 4: Look for an active incident and outage reports
If the status dashboard shows trouble, the next question is scope. Is it a total outage or a partial one? On each service page we break reports down by type: people who can’t log in, people seeing errors, people experiencing slow responses, and people with no access at all. A spike concentrated in “login” problems is very different from a flood of “no access” reports, and it tells you whether a fast fix on your side is even possible.
Outage reports also give you something a status page can’t: a sense of how widespread and how fresh the problem is. When you see a sharp rise in the last few minutes, you’re in the early phase of an incident, so expect a wait. When reports are tapering off, recovery is usually underway.
Step 5: Try the fixes that actually work
If the dashboard says the service itself is fine, these are the moves that resolve the vast majority of AI-tool problems. Work through them in order and stop as soon as something works.
- Hard refresh the page. Reload while bypassing the cache to pull a fresh copy of the interface.
- Open a private window. This skips most extensions and uses a clean session, the single most effective test.
- Sign out and back in. Fixes expired tokens and stuck sessions.
- Clear the site’s cache and cookies. Targeted clearing avoids logging you out of everything else.
- Switch network. Move from Wi-Fi to mobile data (or vice versa) to rule out a local network or DNS issue.
- Try another browser or device. If it works elsewhere, the original environment is the culprit.
- Disable your VPN, then retry the request.
- Wait out a rate limit. If you’re told you’ve hit a limit or the service is “at capacity,” that’s not an outage. Give it a few minutes and try again, or sign in to a paid tier if you have one.
Step 6: When it’s a genuine outage, do this
Sometimes there’s nothing to fix. The service is genuinely down and the only real variable is time. Here’s how to handle a confirmed outage without wasting the next hour refreshing.
- Subscribe to status updates. Most official status pages let you get notified when the incident is resolved, so you don’t have to keep checking.
- Don’t repeatedly resubmit. Hammering a struggling service rarely helps and can cost you work if a half-finished response is lost.
- Switch to an alternative. If you need to keep moving, another AI tool will usually do the job. If your usual assistant is down, a different chatbot can cover most everyday tasks.
- Save your prompt. Copy your last message somewhere safe so you can pick up instantly when the service returns.
How long do AI outages usually last?
There’s no fixed answer, but most outages at the major AI providers are short, often resolved within a few minutes to an hour. Brief slowdowns and “high demand” periods tend to clear the fastest, because they’re capacity issues rather than broken systems. Larger incidents that affect logins or core infrastructure can run longer. The single best predictor is the official status page combined with the direction of outage reports: rising reports mean keep waiting; falling reports mean recovery is close.
Quick links: check any AI tool’s status
| AI tool | Live status |
|---|---|
| ChatGPT | Is ChatGPT down? |
| Claude | Is Claude down? |
| Gemini | Is Gemini down? |
| DeepSeek | Is DeepSeek down? |
| Grok | Is Grok down? |
| Copilot | Is Copilot down? |
| Character.AI | Is Character.AI down? |
Frequently asked questions
How do I know if an AI tool is down or it’s just me?
Check a live status monitor first. If many people are reporting problems and the status is red or amber, it’s the service. If the status is green, test a private window and a different network. The issue is almost certainly local.
Is there one place to check every AI tool at once?
Yes. Our AI status dashboard shows the live status of all major AI tools on a single page, so you can see at a glance whether the problem is one service or a wider outage.
What does “at capacity” or “high demand” mean?
It means the service is temporarily overloaded, not broken. It usually clears within minutes. Waiting a short while, or using a paid tier if you have one, typically gets you back in.
Should I keep refreshing during an outage?
No. Constant refreshing won’t speed up recovery and can lose your in-progress work. Subscribe to the official status updates, switch to an alternative if you need to keep working, and check back once reports start falling.
Next time a tool stalls, start at the is-down.ai dashboard. Confirm first, troubleshoot second, and you’ll spend a lot less time staring at a spinner.