HTTP Latency Test

Measure round-trip latency and jitter from your browser to any HTTPS endpoint - the closest a browser can get to ping.

Live lookup Network Diagnostics Queries a public API
Must be HTTPS. Cross-origin requests are sent opaque - timing still works.
Try:

Result

Ready
Pick a target and press Run. Each request is cache-busted so you measure the network, not your browser cache.

About HTTP Latency Test

Browsers cannot send ICMP, so a real ping is impossible from a web page. What you can measure precisely is HTTPS request round-trip time, which includes the same network path plus TLS and server handling. For comparing endpoints, watching jitter or proving a link is degraded, it is more than good enough.

What is actually being measured

Each sample covers DNS (cached after the first), TCP handshake or connection reuse, TLS resumption, the request, server processing and the response. Connection reuse means later samples are usually much faster than the first - that first sample is your connection setup cost, and it is worth noticing separately.

Latency budgets that matter

Under 40 ms feels instant. Voice stays comfortable to about 150 ms one-way (ITU-T G.114). Interactive SSH becomes annoying past 200 ms. Jitter matters more than absolute latency for real-time media: a stable 120 ms is far better than an average of 60 ms that swings by 80.

Common use cases

  • Comparing latency to several providers or regions from a user location.
  • Demonstrating that a link is jittery when a user reports "the internet is slow".
  • Sanity-checking a CDN or edge deployment is serving from a nearby POP.

Edge cases and gotchas

  • Results include server-side processing time and cannot be compared directly with ICMP ping figures.
  • Cross-origin responses are opaque, so failures cannot be distinguished from CORS rejections - a target that returns all failures may still be up.
  • Browser connection pooling, HTTP/3 and proxies all shift the numbers. Use the same target when comparing over time.

Frequently asked questions

Why is the first sample so much slower?
It includes DNS resolution, the TCP handshake and a full TLS negotiation. Subsequent requests reuse the connection, which is exactly the saving that HTTP keep-alive exists to provide.
Can I test a device on my LAN?
Only if it serves valid HTTPS. Browsers block requests from a secure page to plain HTTP, and self-signed certificates will fail silently.