Perplexity status

How we check Perplexity

We watch api.perplexity.ai/chat/completions from all five of our probe regions, every 2 minutes. A region that sees a failure does not make this page say "down" on its own: a second region has to agree first, the same confirmation rule every YoPingMe monitor uses.

We watch the API host applications call to run Perplexity's search-augmented models, and pin the response an unauthenticated request receives rather than assuming a 2xx. The consumer web app runs behind an anti-bot challenge that an anonymous monitor cannot pass honestly, which would produce false failures unrelated to whether Perplexity is actually up.

What is Perplexity?

Perplexity answers questions by combining a live web search with a language model, which is a different architecture from a model answering purely from what it learned during training. That design is also why availability here means something slightly different than it does for a closed model: a request can depend on the search step, the model step, or both, and either can be the source of a slow or failed response.

Rate limiting is the most common cause of API errors that get mistaken for an outage. Limits are enforced per plan and per model, and a traffic spike or a batch job can exhaust the allowance while the API answers every other caller normally. The response names which limit was hit, which is the fastest way to tell throttling apart from a genuine problem, and it is worth checking before assuming anything is down.

Citation count is a Perplexity-specific detail worth understanding before treating a thin answer as a fault. Because responses are grounded in a live web search, how many sources come back depends on how much relevant material that search actually finds for a given query, which varies naturally with how narrow, recent, or obscure the topic is. A completely healthy API can return a short answer with few citations for an unusual query and a long, well-sourced one for a common query in the very same minute.

The consumer web app and the developer API are separate products built on shared underlying infrastructure, and they can degrade independently of each other. An incident specific to the web app's frontend does not necessarily reach the API, and the reverse holds too, which is why this page watches the API directly rather than the consumer site: that is the dependency an application actually has, distinct from whether the website people browse happens to be reachable at that moment.

Because the underlying web search is live rather than a fixed dataset, identical queries run minutes apart can legitimately return different sources and somewhat different phrasing, the same way a search engine's own results shift over time. That variability is a property of building on live search, not evidence of instability, and it is worth designing around for anything that expects a stable, repeatable answer from the same question asked twice.

Frequently asked questions

Is Perplexity down right now?

Check the board above; it reflects our own probes against the Perplexity API from all five of our regions, refreshed every couple of minutes and confirmed across two regions before this page would call it down.

I am getting 429 errors from the Perplexity API. Is that an outage?

No, a 429 means the request reached the API and was rate limited, not rejected as broken. Limits apply per plan and per model, and a burst of traffic or a batch job can exhaust them while the service answers everyone else normally. The response identifies the limit that was hit, which settles whether you are being throttled or something is actually wrong.

My answer came back without citations or with fewer than usual. Does that mean something is wrong?

Usually not. Citation count depends on how many relevant sources the underlying search found for a given query, which varies with how narrow or recent the topic is, not with the platform's health. A query about a very specific or very new event can return fewer citations on a completely healthy day.

Does a Perplexity outage affect the consumer web app the same way it affects the API?

Not necessarily. The web app and the API are separate products built on shared infrastructure, and one can be degraded without the other. This page watches the API specifically, since that is what a business built on Perplexity actually depends on.

Why did the same query return different sources when I asked it twice?

Because the underlying web search Perplexity performs can return different results between two separate queries, even close together, the same way a search engine's results shift over time. That variability is expected behavior in a live search-backed system, not a sign anything is broken.