Does Incognito Mode Stop Fingerprinting? IP, DNS, WebRTC Compared
Incognito mode does not stop fingerprinting, and it does not hide your IP. What it does is narrow: browsing history, cookies and form data from that window are not kept on your machine after you close it. So when people ask whether incognito mode defeats fingerprinting, start by separating the signals a site uses to recognize you: exit IP, DNS, WebRTC addresses, Canvas and hardware traits, time zone and language. Most of these have nothing to do with whether your machine saved a record.
Below, we take it layer by layer: a summary table first, then a comparison test you can run yourself, then where to go for each layer if you really do want less exposure.
What does incognito mode actually do?
The name differs by browser: Chrome calls it Incognito, Edge calls it InPrivate, Firefox calls it Private Browsing, and Safari calls it Private Browsing too. The core behavior is much the same:
- It opens a separate session. Cookies, localStorage, IndexedDB and cache live in temporary storage, isolated from your normal windows.
- When you close all incognito windows, that temporary data is wiped and no history is written.
- Most browsers don't load extensions in incognito windows by default unless you explicitly allow it.
- Downloaded files and new bookmarks are kept, because they aren't "session data."
Chrome's official help page is blunt about this: incognito mode doesn't hide your activity from the websites you visit, your employer or school, or your internet service provider. Everything below follows from that sentence.
What changes and what doesn't?
This is what should happen, in principle, when you open /en/ip on the same computer and network, first in a normal window and then in an incognito window. Browsers differ in their tracking protection, so for the fingerprint row, trust what you measure yourself.
| Check | Normal → incognito window | Why |
|---|---|---|
| Exit IP (IPv4 / IPv6) | Unchanged | Determined by your system network and routing; the window type has no say |
| IP type, location, reputation | Unchanged | Follows the exit IP |
| DNS resolver | Usually unchanged | Uses the system resolver or the browser's DoH setting, which incognito windows share |
| WebRTC exposed addresses | Unchanged, possibly worse | If you block WebRTC with an extension, incognito windows don't load extensions by default, so the block stops working |
| Canvas / WebGL fingerprint | Unchanged in most cases | Depends on GPU, driver and system font rendering; a few browsers add noise in private windows |
| Hardware fingerprint (core count, memory tier, screen) | Unchanged | Reads parameters of the device itself |
| Time zone, language, fonts | Unchanged | Come from system settings |
| User-Agent, engine version traits | Unchanged | Same browser program |
| Cookies and login state | Changed: starts blank | This is the part incognito mode really isolates |
| Local history and cache | Changed: cleared on close | Same as above |
Only the last two rows change. The rows above them are exactly the parts that can identify you without cookies; What is a browser fingerprint explains them one by one.
How do you run this comparison yourself?
You don't need to install anything. Test once in each window.
- Fix the environment. Don't switch proxies or networks. Keep the same exit for the whole test.
- Test in a normal window. Open cdnping's /en/ip and wait for every check to finish. Write down: exit IPv4, exit IPv6 (if any), the resolver shown by the DNS leak check, the addresses exposed by the WebRTC check, the digests of the hardware and Canvas fingerprints, and your time zone and language.
- Generate a result image. The page can produce a watermarked result image, and you can choose to hide the IP so it's easy to save and compare. Fingerprints are computed locally and not uploaded.
- Repeat steps 2 and 3 in an incognito window. Chrome / Edge use Ctrl+Shift+N, Firefox uses Ctrl+Shift+P, and on macOS replace Ctrl with Cmd; Safari uses Cmd+Shift+N.
- Compare row by row. Fill in both results in this table:
Check Normal window Incognito Same?
Exit IPv4 ____________ ____________ [ ]
Exit IPv6 ____________ ____________ [ ]
DNS resolver ____________ ____________ [ ]
WebRTC addresses ____________ ____________ [ ]
Canvas digest ____________ ____________ [ ]
Hardware digest ____________ ____________ [ ]
Time zone / lang ____________ ____________ [ ]
- Repeat with a second browser (optional). Putting Chrome and Firefox results from the same computer side by side says more than a single browser does: what usually differs is the User-Agent and Canvas, and what stays the same is IP, time zone and hardware.
If in step 5 the incognito window exposes one more WebRTC address, it's almost certainly because an extension isn't enabled in incognito windows.
Why can't incognito mode hide your IP?
The IP lives at the network layer. Every request the browser sends leaves through the system's network stack, passes your router and your ISP, and finally reaches the website. An incognito window only changes where cookies are stored. The request takes exactly the same path, so the source address the site sees is the same too.
Only a change at the network level can change your exit IP: switching networks, or going through a proxy or a corporate egress. Even then it matters whether the rules cover all of the browser's traffic. On a dual-stack network you can end up with IPv4 going through the proxy while IPv6 still connects directly, so a site may see two different exits; see IPv6 proxy not taking effect for how to troubleshoot that. What your IP address reveals covers what can be read from an exit IP: rough location, ISP, data center versus home broadband, and reputation history.
Do DNS and WebRTC behave differently in an incognito window?
DNS: Incognito windows use the same resolver as normal windows. If your system uses your ISP's DNS, the ISP sees every domain you visit, whether or not incognito is on. DNS over HTTPS enabled in the browser is a global setting and applies to incognito windows as well. To change this layer, see the DNS over HTTPS setup guide; the underlying principles are at /en/learn/dns-leak-webrtc-leak.
WebRTC: This is the item people misjudge most often. Many install an extension in their normal window to restrict WebRTC, see no leak in the test, and assume incognito is safe too. In fact, incognito windows in Chrome and Edge disable extensions by default, so that restriction is gone. There are two fixes:
- Turn on "Allow in Incognito" for that extension on the extensions page;
- Or restrict WebRTC through the browser's own settings instead of an extension. See How to disable WebRTC leaks for where the setting is in each browser.
After changing it, go back to /en/ip and test WebRTC again in the incognito window to confirm it took effect.
Why does the fingerprint barely change in an incognito window?
A browser fingerprint is a combination of many traits, and most of them read from the device and the system:
- Canvas / WebGL: The browser draws a graphic and reads the pixels back. GPU model, driver version and font anti-aliasing all affect the result, and none of these change in an incognito window. For the principle and a self-test, see Canvas and WebGL fingerprinting.
- Hardware parameters: Logical CPU core count, memory tier, screen resolution and pixel ratio.
- Time zone, language, font list: Come from the operating system. A time zone that doesn't match your IP location is itself a signal; Browser environment consistency discusses this kind of combined judgment.
Some browsers enable stronger tracking protection in private windows, for example adding random noise to Canvas reads or limiting enumerable fonts. The coverage and default state of these protections change between versions, so go by the browser's official documentation and verify with the comparison steps above rather than trusting marketing text.
One more point that is easy to miss: making your fingerprint more unusual can make you easier to recognize. If only a small fraction of people worldwide run a rare combination, that combination becomes an identifier in itself.
Where is incognito mode useful?
It helps in these situations:
- Shared computers. Logging into email at an internet cafe, on a meeting-room PC or on a friend's computer: after you close the window, no login state or history is left behind.
- Viewing a page as a "new visitor." When debugging a front-end issue, you want to see the first-visit experience with no cookies or cache, or check what a page shows when logged out.
- Isolating two sessions on the same site. For example, you're logged into your own account in a normal window and look at the page's public state in an incognito window, with no cookie crossover. Note that this is about development and debugging, not operating multiple accounts on a platform at once; the latter must follow that service's terms.
- Keeping one search from affecting local recommendations and autocomplete.
For cache-related problems, reproducing in an incognito window is usually easier than clearing the cache by hand.
What does login state in an incognito window leave behind?
An incognito window is not "a new identity for each tab." Within one incognito session, all incognito tabs share the same cookies. If you log in to a site in one incognito tab, another incognito tab opened to the same site is logged in too, until you close every incognito window.
Also, once you log in, the site knows who you are, and nothing you do about your fingerprint or IP matters anymore. The account itself is the strongest identifier.
Which layers do you need to handle separately to reduce tracking?
This checklist lists the measure and the way to verify, layer by layer:
| Layer | Does incognito cover it? | What to do | How to verify |
|---|---|---|---|
| Local records | Yes | Use an incognito window and close all windows when done | Reopen the browser and check history |
| Cross-site cookie tracking | Partly | Turn on the browser's third-party cookie restrictions and tracking protection | Browser privacy settings page |
| Exit IP | No | Handle at the network level; on dual-stack networks check that IPv6 is consistent | /en/ip, compare IPv4 and IPv6 |
| DNS | No | Configure DoH or a trusted resolver | DNS leak check on /en/ip |
| WebRTC | No | Browser setting, or an extension enabled in incognito | WebRTC check on /en/ip, run in an incognito window |
| Fingerprint | Mostly no | Pick a browser or settings with fingerprint protection, and keep a common configuration | Compare both window types and see whether the digests change |
| Account login | No | Don't log in when you don't need to | None |
After making adjustments, you can walk through the developer network self-check list in order, from exit IP all the way to latency characteristics, to make sure no layer was missed.
FAQ
Is InPrivate different from Chrome's Incognito mode? On "no local record," there's no essential difference. The differences are in the default tracking protection level and extension policy, which depend on each version's settings page.
Can my employer or school see what I visit in incognito mode? Yes. Network administrators see the traffic passing through the network and DNS queries, regardless of the browser window type. If management software is installed on a company device, they can see even more.
Can sites still track me with cookies while incognito is on? During that incognito session, yes; cookies work normally within the session. When all incognito windows are closed, those cookies are cleared and the next time starts a new session. But a site may still link the two sessions through your IP and fingerprint.
Why do results on cdnping /ip look almost identical in an incognito window? That's the expected result. The test page reads network and device traits, and incognito mode isn't meant to change those. If there are differences, they usually come from WebRTC (extension not enabled) or Canvas (the browser adding noise in private windows).
cdnping is a free network observatory for developers, with no login and no data sales. You can redo the comparisons in this article at any time on /en/ip; for how the checks work, see /en/methodology.