TL;DR: A kill switch is only real if disconnecting the VPN blocks traffic—or keeps egress on the tunnel interface—instead of silently falling back to your ISP. Capture baseline and VPN IPs with a what is my IP check, then fail the tunnel on purpose and read the result again.
What you are trying to prove
VPN apps advertise kill switches so a dropped tunnel does not expose your real ISP address to the sites you already opened. Proof requires three observations: ISP egress with VPN off, tunnel egress with VPN on, and behavior immediately after a forced disconnect. Icons lying about “protected” do not count.
Privacy goals differ from corporate compliance goals, but both need the same evidence trail. A screenshot of the vendor marketing page is not evidence; before/after public IPs are.
Establish baselines before chaos
On a quiet network:
- Turn VPN off. Copy the public IP and ASN from www.whatismyipaddressinquire.com.
- Connect VPN to a named server. Confirm the public IP changed and note the new ASN.
- Save both strings with timestamps in a scratch note.
If the IP never changes, fix the VPN path before testing kill-switch myths. Also note whether IPv6 addresses appear; dual-stack surprises belong in the baseline, not as a postmortem discovery.
Force a disconnect the way reality does
Do not only click Disconnect in the app (that may disable the kill switch by design). Prefer: disable the virtual adapter, block the VPN protocol in a local firewall for a minute, or yank the underlay network briefly and restore it. Watch whether browsing still works and which public IP appears.
Outcomes:
- Hard fail closed: pages time out; IP check fails—kill switch likely holding.
- Fail open: pages load on the original ISP IP—kill switch absent or bypassed.
- Unexpected third IP: another agent (corporate proxy, private relay) took over—investigate before trusting either reading.
DNS path is a separate assertion
Even when egress IP stays on-tunnel, DNS might query the ISP resolver. Pair your IP check with a resolver identity test your vendor documents. Kill-switch marketing that ignores DNS still leaks browsing metadata. If your threat model cares about ISP-visible queries, treat DNS fail-open as a failed test even when the public IP stays perfect.
IPv6 leak patterns
Some clients secure IPv4 only. If your ISP offers v6, a dual-stack site may see the native v6 address while v4 looks tunneled. Disable v6 temporarily for the test or use a lookup path that shows both families. A clean v4 kill switch with a naked v6 address is still a fail for privacy goals.
Re-enable v6 only after the client demonstrates that v6 also dies or stays tunneled during a forced disconnect. Leaving v6 disabled forever is a blunt workaround, not a completed test.
App vs OS vs router enforcement
Application kill switches can be bypassed by apps that bind funny interfaces. OS-level firewall rules or router-enforced VPN egress are stricter. Know which layer you bought. Test the layer you actually rely on—do not assume the consumer app toggle equals network-wide policy.
Travel routers that force all LAN clients through a VPN deserve their own kill-switch drill: pull the upstream WAN while a phone on Wi-Fi tries to load a page. The phone should stall, not roam to cellular mid-test unless you intentionally allow that path.
Split tunneling complicates the story
If your profile excludes banking or LAN destinations from the tunnel, those destinations will always show ISP egress by design. Do not mark the kill switch “failed” when an excluded route uses the ISP—document exclusions first. Kill-switch tests target traffic that was supposed to stay inside the tunnel.
Write a tiny retest ritual
After every VPN app update, repeat baseline → connect → forced fail → observe. Takes five minutes. Updates have shipped with broken firewall hooks before. Teams that skip retests discover fail-open behavior during travel week, not during a calm Tuesday. Store the ritual in the same doc as your corporate VPN install guide.
Interpreting results without drama
If the switch fails open, stop sensitive browsing on that config, enable a stricter mode if offered, or switch products. Keep the IP evidence; vendor support responds better to before/after addresses than to “I feel exposed.” A public IP check cannot configure your VPN—but it can prove whether the kill switch earns its name.


