Is ChatGPT down right now?
This page shows the current operational status of ChatGPT, OpenAI's chat assistant. The status indicator at the top is updated every 60 seconds by polling OpenAI's official status feed and combining it with real-time user reports. If the indicator is green, the public ChatGPT API and web app are both responding normally. If it is amber, one or more components are degraded — typically slow response times or partial regional outages. A red indicator means a major incident is in progress and a meaningful number of users cannot use the service.
Because companies often lag thirty to forty-five minutes behind reality on their official status pages, we surface user-submitted reports alongside the official feed. If you are seeing errors but the dot is still green, scan the report counter — a sudden spike usually beats OpenAI's announcement by half an hour.
ChatGPT outage history and common causes
ChatGPT is a complex stack with multiple layers that can fail independently. The model serving infrastructure can run out of capacity during peak hours, especially in the hours after major product announcements. The web frontend at chat.openai.com sits on its own deployment pipeline and can be broken by a bad release while the underlying API is fine. The authentication layer — built on Auth0 — has caused full lockouts in the past even when the rest of the system was healthy.
Common causes of ChatGPT outages over the past year include: capacity overloads during global traffic spikes, DDoS-style attacks that have triggered rate-limiting cascades, upstream provider issues at Cloudflare and AWS, expired SSL certificates, and bad model deployments that cause specific features (Code Interpreter, Vision, GPTs) to fail without taking down the whole product. Plus and Pro subscribers occasionally see issues that free users do not, because paid tiers run on partially separate infrastructure.
If you want to understand the pattern: short outages of five to fifteen minutes happen roughly once a week and are usually invisible to most users. Major incidents lasting an hour or more happen roughly once a month and tend to cluster around new launches.
Is ChatGPT reliable?
For an under-five-year-old consumer product running at hundreds of millions of weekly users, ChatGPT's uptime is reasonable but not exceptional. Independent monitoring puts its 30-day availability somewhere in the 99.0 to 99.5 percent range, depending on which features you count. That is materially worse than mature SaaS infrastructure (Slack, Gmail, Stripe), which typically sits above 99.9 percent. The headline numbers also hide a meaningful tail of partial outages where ChatGPT loads but specific tools (Vision, file uploads, GPTs, voice mode) silently fail.
For mission-critical use, the practical advice is simple: do not depend on a single provider. The major chat assistants — ChatGPT, Claude, Gemini — rarely go down simultaneously. Most professional users keep at least two open. If your workflow truly cannot tolerate downtime, the OpenAI API has somewhat better uptime than the consumer product because it does not share the same web frontend stack.
What to do when ChatGPT is down
First, confirm it is actually a ChatGPT problem and not your network. The status indicator at the top of this page is the fastest signal. If it is red or amber, the issue is on OpenAI's side and refreshing will not help.
Second, try a different entry point. The OpenAI API often works when the web app does not. The mobile apps occasionally work when the desktop site is broken, and vice versa. Logging out and back in resolves authentication-tier issues but does nothing for capacity issues.
Third, switch tools. Claude and Gemini can both handle most of what users typically do in ChatGPT — writing, coding, research, analysis. Their status pages are also linked from this site so you can check them in one click. Perplexity is a strong alternative for research and search-augmented questions specifically.
Finally, if you are blocked on something time-sensitive, report the outage on this page. The user report counter helps everyone else trying to figure out whether to wait or move on, and gives us a faster signal than OpenAI's own status page.