← All articles

undetected-chromedriver and nodriver, what they fix and where they stop

4 min read seleniumautomationguide

You moved a Selenium scraper to undetected-chromedriver because a site kept flagging ChromeDriver, and it worked. Months later, with more accounts on the same machine, the flags came back, and switching to nodriver did not change that. Nothing broke. You ran into the edge of what both projects were built to do.

What undetected-chromedriver changes

undetected-chromedriver describes itself as an optimized Selenium Chromedriver patch that automatically downloads the driver binary and patches it. The target is the driver process, the thing that injects recognisable objects into every page it controls. Its README notes that the approach moved from removing and renaming those injected variables to preventing them from being injected at all.

The practical appeal is that your code stays Selenium. uc.Chrome() returns a WebDriver, your locators, waits and page objects keep working, and the change is a few lines. For a Selenium codebase that is the cheapest possible fix for a driver-level tell.

The README is also direct about the limits. It says, in capitals, that the package does not hide your IP address, that running from a datacenter makes it likely you will not pass, and that headless mode is not officially supported.

What nodriver changes

nodriver is presented as the official successor. It removes the driver instead of patching it: no chromedriver binary, no Selenium dependency, and the library talks to Chrome directly over the DevTools protocol. It is fully asynchronous, so it is a rewrite rather than an import swap. Its stated goal is to stay undetected for most anti-bot solutions, and it does not promise more than that.

Two details in its README matter for account work. By default it uses a fresh profile on each run and cleans it up on exit, so cookies and storage are gone unless you pass your own profile directory. And its expert mode, which loosens web security, is described by the README itself as making you more detectable.

The common boundary

Both projects answer one question: is this browser being driven by automation? They take away the evidence that says yes. They do not touch the question a risk engine asks next: which device is this?

Canvas output, WebGL renderer strings, installed fonts, audio processing, screen and hardware values all come from the real machine and the real Chrome install. Patching the driver does not change them. So:

This is not a criticism. Removing the automation tell is a real and necessary step, and these projects do it for free. Device identity is a different problem that they never set out to solve.

When they are enough

They are the right answer, and switching would be overhead, when:

If that describes you, stay where you are. Spend the effort on proxy quality and pacing instead.

When you need something else

Move on when several accounts on one machine must not share a device, when a profile has to survive restarts and Chrome updates with the same identity, or when proxy, timezone and locale have to agree without hand-written glue.

AntiBrow generates a fingerprint once per profile, applies it in the browser engine below the JS layer, and persists it with the profile. Binding a proxy makes timezone, locale and geolocation follow the exit IP, and profiles are directories on your own disk, unlimited on every plan including Free.

The cost is real. launch() returns a Playwright Page and BrowserContext, not a Selenium WebDriver and not a nodriver tab. Coming from undetected-chromedriver means rewriting locators, waits and any Grid setup. Coming from nodriver means mapping its async API onto Playwright's, which is closer but still a rewrite. The driver-level leaks themselves are covered in Selenium automation leaks; this post is about what remains after you fix them.

Check which problem you have

  1. Open three profiles with your current setup on the same machine. Load a canvas test such as the one described here in each and compare hashes. Identical hashes mean your accounts share a device.
  2. In each, compare Intl.DateTimeFormat().resolvedOptions().timeZone with the exit IP's location.
  3. Read navigator.webdriver in the console. If it is false and steps 1 and 2 are clean, the automation layer is not your problem and neither is identity. Look at IP reputation and behaviour next.
  4. If step 1 failed, no driver patch will fix it. Try one free AntiBrow profile with the same check, following the SDK docs.

Try it on the free tier.

Unlimited local profiles, no credit card. Check it against the detectors yourself.