Perplexity Not Working?
Live status check + step-by-step troubleshooting for Perplexity AI outages, failed searches, and API errors.
Live Perplexity Status
Updated every 5 minutes → prismix.dev/service/perplexity If this badge shows Operational but searches still fail, the issue is likely a specific model mode or your account quota — not the platform itself.
What's Wrong? Diagnose Fast
Search returning empty or wrong answers
Try switching Focus mode (Web — Academic — Reddit) or change the model to default Sonar. May be a model-specific degradation.
Searches loading forever or timing out
Web search backend or the model synthesizer is overloaded. Check status.perplexity.ai. Try a simpler query without Citations mode.
Pro features locked or quota error
Pro gives 300+ advanced searches/day. If you hit the limit, basic Sonar searches still work. Check perplexity.ai/settings for usage.
API returning 429 or 401 errors
429 = rate limit hit. 401 = invalid API key. Check your key at perplexity.ai/settings/api and add exponential backoff for 429s.
Step-by-Step Fix
- 1
Check live status first
Visit prismix.dev/service/perplexity or status.perplexity.ai. If there's an active incident, wait for resolution — no local fix will help.
- 2
Identify which stack is failing
Perplexity has three independent components: the web app (perplexity.ai), the API (api.perplexity.ai), and the model backends (Sonar, GPT-4o Sonar, Claude Sonar). Each can fail independently. Try the web app if you normally use the API, or vice versa.
- 3
Switch models or Focus mode
From the search bar, click the model selector and switch between: Default Sonar, Claude Sonar, GPT-4o Sonar. If one model is degraded, another usually works. Also try a different Focus (Web, Academic, YouTube, Reddit, Writing).
- 4
Disable extensions and clear cache
Open perplexity.ai in a fresh incognito window with all extensions disabled. uBlock Origin and some privacy extensions block Perplexity's WebSocket connections causing silent failures.
- 5
For API: add backoff and check key validity
For 429 errors, implement exponential backoff starting at 1 second, doubling up to 5 retries. For 401, regenerate your key at perplexity.ai/settings/api. Free tier is limited to 5 requests/minute; Pro tier is 60/minute.
Perplexity Models and What's Available
| Model | Available on | Best for |
|---|---|---|
| Sonar (default) | Free + Pro | General web search, fast |
| Sonar Pro | Pro only | Deeper research, multi-step queries |
| GPT-4o Sonar | Pro only | Coding, structured output |
| Claude Sonar | Pro only | Long documents, nuanced writing |
| Sonar Reasoning | Pro only | Step-by-step reasoning tasks |
API Rate Limits
| Tier | Requests/min | Monthly cost |
|---|---|---|
| Free | 5 | Pay-per-token ($0.20–$1/M tokens) |
| Pro | 60 | $20/mo (includes 300+ web searches) |
| Enterprise | Custom | Contact sales |
Get Alerted When Perplexity Goes Down
Star Perplexity on Prismix and receive email or webhook alerts the moment status changes — before the Twitter complaints start.
View Perplexity Status →Frequently Asked Questions
Why is Perplexity not working?
Common causes: platform outage (check prismix.dev/service/perplexity), Pro quota exhausted, a specific model (Claude Sonar/GPT-4o Sonar) is degraded, or a browser extension blocking perplexity.ai scripts.
Is Perplexity down right now?
Check prismix.dev/service/perplexity for live status updated every 5 minutes, or status.perplexity.ai directly. If the platform is operational but searches fail, try switching the model or Focus mode.
Perplexity Pro not searching — what to do?
Pro gives 300+ advanced searches/day. If you hit the limit, basic Sonar searches still work. Try disabling extensions, switching models (GPT-4o Sonar vs Claude Sonar), or refreshing with cleared cache.
Perplexity API 429 — how to fix?
Rate limit hit. Free tier: 5 req/min. Pro: 60 req/min. Add exponential backoff (1s, 2s, 4s up to 5 retries) and cache results where possible. Upgrade to Pro for higher limits.
Search results are empty or clearly wrong — why?
Try switching Focus (Web/Academic/Reddit), change the model to default Sonar, or disable the "Citations" requirement. Some query structures confuse the search-then-synthesize pipeline.