About Is AI Down
A small, independent project that turns anonymous crowd reports into readable status for major AI products—without pretending we are inside OpenAI's or Google's NOC.
Why it exists
When ChatGPT throttles, Claude errors, or an API times out, the first question is usually "is it just me?" Vendor status pages are not always timely, and social feeds are loud. Is AI Down is meant as a third place: structured reports, rolling windows, and a stable headline you can skim in a few seconds.
We do not replace official incident communications. We mirror what people are experiencing in the open, with math that down-weights one-off noise and up-weights coordinated spikes.
What makes it different
Crowd-sourced signal
Reports come from people using ChatGPT, Claude, Gemini, and the rest in the wild—not from a paid panel. That keeps the signal grounded in real failures, slow responses, and odd behavior you actually see.
Time windows, not vibes
We score activity across short, medium, and long lookbacks so a single angry click does not equal an outage, but a coordinated spike in a few minutes does.
Public methodology
The rules live in plain language on the site. Every headline status ships with a confidence hint so you know when we are sure versus when the data is still thin.
Live pages
When new reports land, calculations rerun and open dashboards can refresh over WebSockets—you are not waiting for a nightly batch file to learn that something broke.
How status is built
Inputs
- •Reports: categorized issues (timeouts, bad output, auth, billing, and similar) submitted from each platform's page.
- •Windows: the backend compares volume to a rolling baseline across overlapping slices—on the order of minutes, an hour, and a day.
- •Separation: a lone report should not look like a continent-wide outage; a synchronized burst should.
Labels you will see
Older cards may still show legacy wording (for example "Operational" or "Minor issues"). New work standardizes on the five below.
- Likely Operational: Report volume looks quiet compared with what we expect; nothing suggests a widespread issue.
- Possible Issues: Enough fresh friction that something may be wrong, but the pattern is still noisy or early.
- Potential Problems: Stronger uptick or mixed signals—worth watching before calling it a full incident.
- Likely Down: A tight burst of aligned reports—the sort of shape you see when many people hit the same failure.
- Monitoring Reports: Reports exist but clustering looks localized or inconsistent; we keep the headline cautious.
Full detail: How we calculate status.
Who builds it
I'm Rahul—a solo developer based in India. I use these tools daily and got tired of refreshing status blogs that update slowly. Is AI Down is my attempt to give everyone a shared, honest signal built from the same messy evidence we all already have: what broke for real users, when.
Say hello
Want a new provider on the board, or spotted a logic bug? I read mail even when I cannot ship a fix the same week.
Contact