If you use split tunnelling and want to know whether each app takes the route you chose, a short test beats a connection icon alone, and Surfshark is our recommendation for Pakistan. You can Claim 80% off with Surfshark through our partner link, which is 80% off for our users, then work through the checklist below.
Split tunnelling lets you send some apps or sites through the VPN while other apps or sites use your normal connection. That split can be useful when one service works better on a local connection while your browser or work app needs VPN protection.
It can also cause confusion, because a correct split looks like a leak if you test the wrong app, and a real leak looks like a split if you never test both sides.
This checklist therefore tests both sides of the split in the same session. You will record a baseline, test an app inside the tunnel, test an app outside the tunnel, check DNS, and check what happens when the VPN drops.
Each test uses a simple IP or DNS check, so you do not need specialist tools to get a useful result, although a careful note of each result is essential.
What Split Tunnelling Controls
Split tunnelling controls routing, not encryption for every app at once. Apps you include in the tunnel send their traffic through the VPN server, while apps you exclude send their traffic through your normal internet provider.
The VPN app settings decide which list applies, and the mode matters: some apps use an include list, where only listed apps use the tunnel, while others use a bypass list, where listed apps stay outside it.
That mode difference is the first thing to confirm before you test. If you think you excluded a browser but the app actually uses an include list, the browser may be inside the tunnel by default, or outside it by default, depending on the design.
Read the mode label in your VPN app settings, write down which apps are listed, and test against that written list rather than against a guess from memory.
Split tunnelling also interacts with DNS, the kill switch, and browser extensions. A browser extension VPN only protects that browser, while a full-device app can protect many apps at once. If you mix an extension and a full app, a test in one browser may not describe the behaviour of another app, so test each app you care about separately and note the app name with every result.
Finally, split tunnelling does not change the privacy of excluded traffic. An excluded app uses your normal connection, so your internet provider and the sites it visits can see that traffic as usual. Only exclude apps you can name and justify, such as a local printer tool or a clearly non-sensitive app, and keep sensitive browsing and work apps inside the tunnel by default.
Prerequisites and Safe Test Conditions
Before you start, update the VPN app to the latest version and restart the device if an update was just installed. Old app versions can behave differently after a system update, and a fresh restart removes stale routes that might confuse your first test. Note the app version at the start of your notes, because support will ask for it if a result differs from the expected split.
Choose a safe test environment. Use your own device, your own accounts, and a network you trust for the first test, rather than a shared or public network with extra controls. Public networks can add their own routing, captive portals, or blocks, and those extra layers make it harder to tell whether an odd result came from split tunnelling or from the network itself.
Decide the test pair in advance. Pick one app that should stay inside the tunnel and one app that should stay outside it, and use IP check sites inside both apps where possible. Two browsers are a simple pair for a first test: one browser inside the tunnel and a second browser outside it, provided your VPN app supports per-app routing on your platform.
Close unrelated apps that might use the network during the test, and do not change settings halfway through a test run. A download, a cloud sync, or a background update can change the network state and make a result hard to reproduce. A quiet device, a written app list, and a fixed server choice give you a clean baseline for every check that follows.
- Update the VPN app and note the app version before you test.
- Pick one inside-tunnel app and one outside-tunnel app in advance.
- Use your own device and a trusted network for the first test.
- Close background downloads and cloud sync during the test run.
Record a Baseline Before You Test
A baseline is a short record of your connection before the VPN is involved. With the VPN disconnected, open an IP check site in each test app and note the IP address, country, and provider name shown. Those baseline values are the reference for the outside-tunnel result, because an excluded app should return to a result close to this baseline when the VPN is connected.
Also note the time, the network name or type, and the VPN server you plan to use. A baseline without a time is hard to compare later, because IP addresses can change on some home networks, and a server change can change the expected inside-tunnel result. A few lines of notes at the start prevent a long debate about what changed later in the session.
If you use a DNS leak test later, record a baseline for that too. With the VPN off, note the DNS provider shown by the leak test, then compare it with the result inside and outside the tunnel.
A DNS baseline helps you spot a case where the IP address looks correct but the name lookups still go to your original provider, which is a different kind of failure.
Keep the baseline in the same note as the later results, not in a separate message or screenshot that you might lose. A single test note with baseline, inside result, outside result, DNS result, and kill-switch result is easy to send to support, and it is easy to repeat on another day when you want to check whether an app update changed the split.
Test One: IP Check Inside Versus Outside the Tunnel
Connect to the VPN with your chosen split settings in place, then open the IP check site in the inside-tunnel app first. The expected result is the VPN server IP address and the server country, not your baseline address. If the inside app still shows the baseline address, the app may be excluded by mistake, or the split setting may not have applied until the app restarted.
Next, open the same IP check site in the outside-tunnel app. The expected result is close to your baseline: your normal provider address and your normal country. If the outside app shows the VPN server address instead, the exclusion may not be active, the app may have been restarted into the tunnel, or the mode label may mean the opposite of what you assumed.
Repeat both checks after fully closing and reopening each test app. Some apps only pick up a new route when they restart, and a tab left open from before the VPN connection can show a cached result. A fresh app launch, a fresh tab, and a fresh IP check give a much stronger result than a refreshed old tab that may still hold yesterday network state.
If both apps show the VPN address, or both show the baseline address, the split is not working as listed. Do not assume a bug at once: re-read the mode label, confirm the app names in the list, and check whether a browser extension or a second VPN app is overriding the full-device setting.
Only after those checks should you treat the result as a support case with evidence attached.
Test Two: Per-App Routing Check
A per-app routing check extends the first test beyond browsers. Where your VPN app supports it, add a specific desktop or mobile app to the inside or outside list, restart that app, and then use its own network behaviour or an IP check inside it to judge the route.
For a browser, an IP check site is direct evidence, while for another app you may need its connection log or a visible server region to judge the route.
Test one app at a time. Adding five apps to a list and testing them all at once makes it hard to tell which app behaved differently and why. Change one app, restart it, test it, note the result, and only then move to the next app.
That slow pace feels tedious, but it produces a test note that support can actually follow without asking for a repeat run.
Watch for apps that use helper processes, embedded browsers, or background services. A main app may sit inside the tunnel while its updater or helper uses a different route, and a single IP check inside the main window may miss that helper traffic.
If an app has a separate updater or login helper, note that behaviour in your test note rather than claiming a clean result for the whole app.
Also test after an app update, because updates can reset split settings, rename processes, or change how an app connects. A split that passed last month can fail after an update without any change on your side, so a short retest after major updates is a sensible habit. Note the app version with each retest, so a version change is visible in your history.
Test Three: DNS Check
A DNS check tests name lookups, not just IP addresses. Open a DNS leak test in the inside-tunnel app and note the DNS provider shown. The expected result is the VPN provider DNS or a DNS service you deliberately chose for the tunnel, not your home internet provider.
If the inside app shows your home provider, the tunnel may be carrying web traffic while DNS lookups still take a direct route.
Repeat the DNS check in the outside-tunnel app. The expected result there is your normal DNS provider, because an excluded app should behave like a normal connection. If the outside app shows VPN DNS while its IP check shows a normal address, you have a mixed result that needs a note, because IP and DNS routes can differ when a system setting overrides one of them.
If a DNS result looks wrong, check the VPN app DNS setting first, then check any manual DNS setting on the device. A manual DNS address set at system level can override the VPN DNS for some apps, and a browser DNS setting can override it for that browser alone.
Change one DNS setting at a time, retest, and note which change moved the result toward the expected value.
Test IPv4 and IPv6 separately where the test tool offers that choice. Some systems have IPv6 while the VPN only tunnels IPv4, and a DNS or IP result on IPv6 can reveal a route that an IPv4-only check would miss. If your device has no working IPv6, note that IPv6 could not be tested, rather than treating a missing IPv6 result as a leak or as a pass.
Test Four: Kill-Switch Interaction
The kill switch and split tunnelling interact in ways that vary by provider, so test that interaction deliberately rather than assuming it. With split settings in place and the VPN connected, confirm the inside and outside results first, then force a VPN drop in a safe way, such as switching the server or disconnecting as the provider guides you for a kill-switch test.
Note what happens to each test app during the drop.
In many designs, the kill switch blocks inside-tunnel traffic when the VPN drops, so the inside app should lose internet access until the VPN reconnects. The outside app, by design, may keep working on the normal connection, because it was never supposed to use the tunnel. That mixed behaviour can be correct, but only if it matches the provider documentation for your app version.
If the inside app keeps working on your normal connection during a VPN drop, the kill switch may be off, it may be disabled while split tunnelling is active, or it may apply only to certain apps. Check the kill-switch setting, read the provider help page for split tunnelling and kill-switch interaction, and note the exact setting name.
Do not assume protection that the documentation says is limited in split mode.
Reconnect the VPN after the kill-switch test, then repeat the inside and outside IP checks. A kill-switch test can leave an app in a stale state, and a fresh check confirms that the split returned to normal after reconnection. If an app still shows the wrong route after reconnection, restart the app, recheck, and add that restart step to your evidence note.

Expected Versus Unexpected Results
A results table helps you judge a test at a glance, because it separates a correct split from a leak and from a mislabelled setting. Use the table below as a template for your own note, and fill in the actual addresses and provider names you observed, rather than copying the example labels.
Your own observed values are the evidence, while the table labels only describe what each pattern means.
- Inside app shows VPN server IP: expected for a tunnelled app.
- Outside app shows baseline IP: expected for an excluded app.
- Inside app shows baseline IP: unexpected, check the app list and mode.
- Outside app shows VPN server IP: unexpected, check the exclusion list.
| Test App | Expected Result | Unexpected Result |
|---|---|---|
| Inside-tunnel app | VPN server IP and server country | Baseline IP or home provider DNS |
| Outside-tunnel app | Baseline IP and normal DNS provider | VPN server IP when exclusion is active |
| DNS check inside tunnel | VPN DNS or chosen tunnel DNS | Home provider DNS for tunnelled lookups |
| Kill-switch drop | Inside app blocked until reconnect | Inside app continues on normal connection |
An unexpected result is not always a security failure. A mislabelled mode, an app that needs a restart, or a cached tab can all produce an unexpected pattern without any leak in the VPN itself. The useful response is to re-read the setting, restart the app, retest, and only then decide whether the pattern persists across a clean repeat run.
A persistent unexpected result across a clean repeat run is stronger evidence of a real problem. At that point, keep the settings unchanged, gather the evidence described in the next section, and contact support with the full note. Changing settings repeatedly before you contact support can erase the conditions that produced the problem, and support may then be unable to reproduce it.
Record Reproducible Evidence for Support
Good evidence lets another person repeat your test and get the same pattern. Start with the basics: device type, operating system version, VPN app version, test app names and versions, VPN server country, and the time of each test. Without those details, a support agent cannot tell whether your result came from an old app version, a different server, or a platform limit.
Next, record the observed values in plain text as well as in images. Write down the baseline IP country and provider, the inside result, the outside result, and the DNS provider for each app. Plain text values are searchable and easy to quote back, while images of your own results add visual confirmation of the page and the time shown on screen.
Capture images of your own test results only, showing the IP check or DNS check page, the app name, and a visible timestamp where possible. Avoid including passwords, full IP addresses in a public post, or personal data that support does not need.
For a private support ticket, a full result is useful, but for a public forum post, redact the full address and keep only the country and provider name.
Describe the steps in order, including the split mode label, the apps listed, and whether each app was restarted before the test. A step list such as connected, opened inside app, checked IP, opened outside app, checked IP, then tested DNS is far more useful than a summary that only says the split did not work.
Support can follow a step list and spot a missed restart or a wrong mode label quickly.
Finally, state the expected result and the observed result side by side, and attach the note as one document rather than as scattered messages. A clear statement such as expected inside app to show VPN server country, observed baseline country, is easier to act on than a long narrative.
Keep the original note after you send it, in case support asks for a repeat run with a different server.
Troubleshooting Common Test Failures
If the inside app shows the baseline address, first confirm the app is actually listed for the tunnel and that the mode label matches your intention. Then fully close and reopen the app, because many apps only adopt a new route on launch. If the result still fails, restart the device, reconnect, and retest, since a stale route can survive an app restart on some systems.
If the outside app shows the VPN address, confirm the exclusion list and the mode label in the same way, then restart the outside app and retest. A browser extension, a second VPN app, or a system proxy can also force an excluded app into the tunnel, so check for other network tools before you conclude that the split setting itself is broken.
If DNS shows the wrong provider, check the VPN DNS setting, the device DNS setting, and the browser DNS setting in that order. A single manual DNS address can explain a DNS result that contradicts an otherwise correct IP result, and fixing the DNS setting often resolves the mismatch without changing the split list at all.
If the kill switch does not block the inside app during a drop, read the provider documentation for your exact app version and platform. Some providers limit the kill switch while split tunnelling is active, and that limit is a documented behaviour rather than a bug. If the documentation promises blocking and the test shows otherwise, that gap is exactly the evidence to send to support.
If results change after an update, note the old and new app versions and repeat the full checklist. Updates can reset settings, change mode labels, or alter process names, and a full repeat test is the fastest way to find which setting changed. Do not restore an old version as a fix without checking the provider guidance, because old versions can carry security issues of their own.
Platform Limits and Policy-Only Features
Platform limits shape what split tunnelling can do, and iOS is the clearest example. App-based split tunnelling is unavailable on iOS in the usual per-app sense, because iOS does not give VPN apps the same per-app routing control as Android or Windows.
On iPhone, you may find website or domain bypass options in some apps, but you should not expect the same per-app list you would test on Android or Windows.
That limit means a test plan must match the platform. A checklist written for Windows will not transfer directly to iPhone, and a failed iPhone test may reflect a platform limit rather than a broken setting.
Check the provider help pages for your platform before you test, and note the platform in your evidence, so support does not ask you to repeat a test that the platform cannot perform.
Some features are policy-only, meaning they are described in documentation or settings rather than proven by a single IP check. Examples include a no-logs claim, an audit claim, or a promise about how the provider handles a dropped connection in a rare edge case.
Check those features by reading the current documentation, the settings labels, and any audit summary the provider publishes, rather than inventing a test that cannot prove them.
Do not invent proof for a policy-only feature. If a claim cannot be tested with an IP, DNS, or kill-switch check, say so in your note and cite the documentation page you read, with the date you read it. Honest limits make your other test results more credible, because a reader can tell which claims rest on observed checks and which rest on provider documentation alone.
For readers in Pakistan, platform limits also affect device choice for split tunnelling. If per-app routing is a daily need, Android or Windows gives you more direct controls to test, while iPhone use may need a different workflow, such as a browser-based bypass or a router-level split. Choose the workflow that matches your platform, then test that workflow with the same checklist discipline.
Why Surfshark Is Our Recommended VPN for Pakistan
Surfshark fits this checklist because its apps on Android and Windows offer a bypass feature for apps, websites, and addresses, which gives you a clear list to test against. A testable list matters more than a long feature page, because you can only verify a split when you know exactly which apps the provider says are inside or outside the tunnel.
Surfshark also provides the companion controls this checklist uses, including a kill switch and DNS leak protection, so you can test the kill-switch interaction and the DNS result in the same session. Testing those controls together reveals interactions that a single IP check would miss, such as a kill switch that behaves differently while a bypass list is active.
For readers in Pakistan, unlimited simultaneous connections are again practical, because you can run the same split test on a phone and a laptop without signing one device out to test the other. That convenience does not replace the checklist, and you should still record a separate baseline and result for each device, because a split can pass on one platform and fail on another.
Surfshark — our recommended VPN for Pakistan
Testable bypass lists for apps and sites, plus kill-switch and DNS controls to check.

Frequently Asked Questions
How Do I Test VPN Split Tunneling?
Start with a baseline: with the VPN off, note the IP country and provider in each test app. Then turn the VPN on with your split list in place, and check the same IP tool in an inside-tunnel app and in an outside-tunnel app, restarting each app first so it picks up the new route.
Next, run a DNS check in both apps and note the DNS provider for each one. The inside app should show VPN DNS or your chosen tunnel DNS, while the outside app should show your normal DNS provider, matching the baseline you recorded at the start.
Finally, test the kill-switch interaction by forcing a safe VPN drop as the provider guides you, then reconnect and repeat the inside and outside checks. A complete test therefore covers IP, DNS, and kill-switch behaviour, not just a single IP page.
What Result Should I Expect?
For an inside-tunnel app, expect the VPN server IP address and server country, with VPN DNS or your chosen tunnel DNS. For an outside-tunnel app, expect your baseline IP address and normal DNS provider, close to the result you recorded before you connected.
During a VPN drop, expect the inside app to lose internet access until reconnection if the kill switch covers it, while the outside app may keep working on the normal connection by design. That mixed result can be correct, provided it matches the provider documentation for your platform.
If both apps show the same address, or the DNS provider contradicts the IP result, treat that pattern as unexpected. Restart the apps, re-read the mode label, retest cleanly, and only then treat a persistent pattern as a support case.
What Evidence Should I Send Support If the Result Differs?
Send a single test note with device type, operating system version, VPN app version, test app versions, server country, and the time of each test. Add the baseline result, the inside result, the outside result, and the DNS result, with the split mode label and the exact apps listed.
Add images of your own test results showing the IP or DNS page, the app name, and a timestamp where possible, and describe the steps in order, including whether each app was restarted. Plain text values plus images give support both a searchable record and visual confirmation.
State the expected result and the observed result side by side, and keep the original note for a repeat run. A clear comparison, such as expected VPN server country and observed baseline country, helps support reproduce the issue without several rounds of follow-up questions.
Surfshark — our recommended VPN for Pakistan
Run the full IP, DNS, and kill-switch checklist before you trust a new split list.
Final Verdict
A split tunnelling test is useful only when it checks both sides of the split in the same session. Record a baseline, test an inside app and an outside app, check DNS, and test the kill-switch interaction, then write down the observed values with app versions and times. That discipline turns a vague feeling about routing into a result you can repeat and defend.
Remember the limits in this checklist as well. iOS does not offer the same per-app split tunnelling as Android or Windows, some features are policy-only and must be checked in documentation rather than invented tests, and an unexpected result is often a mode label or a missed restart rather than a broken VPN. Checking those limits first saves time and avoids false alarms.
Use this checklist after every major app update, every split list change, and every move to a new device. A short repeat test on a new setup is far cheaper than discovering a routing mistake during sensitive work, and a clean evidence note makes any support exchange faster for both sides.



