Browser extensions are a fingerprint surface

Adding an extension to an automated browser feels like adding a feature. It is also adding something a page can look for, in three different ways, and the most common reason people add one is to hide a fingerprint, which is where this gets circular.

The three ways a page finds an extension

Web-accessible resources. An extension that exposes files to pages does so at a URL containing its own identifier. Fetch that URL and see whether it loads.

fetch('moz-extension://<id>/<file>')       // resolves only if that extension is here

The identifier differs per browser and, in Firefox, per profile, which limits the scan. It does not eliminate it, because extensions that need page access often need predictable resources.

What it does to the page. An ad blocker removes elements, so a page can insert a bait element with a name matching a common filter list and check whether it survived. A style-injecting extension changes computed values. Anything that modifies the DOM leaves the modification behind, and the modification is the detection.

The behaviour it adds. An extension that patches navigator or the canvas has to run before the page, and it leaves the same traces as any other page-level override: own properties where prototypes are expected, getters that are not native, a toString that prints your source.

Why a stealth extension is usually the wrong shape

This is worth stating plainly, because it is the most common reason someone asks how to load one.

A stealth extension operates at the page layer. It runs inside the document, it is made of JavaScript, and everything on the three levels page about that layer applies to it. It also has to run before the page’s own scripts, which is a race it does not always win.

And if the browser already sets those values in its engine, the extension is now a second opinion arguing with the first. Two components deciding what navigator says do not average out; they disagree, and the disagreement is easier to spot than either answer alone. That is the same argument as turning off a crawler’s own fingerprint generator.

So: extensions for functionality, yes. Extensions for stealth on top of a stealth engine, no.

Extensions that are fine, and how to think about the rest

Ask what the extension does to the page:

Nothing visible to the page. A tab manager, something that only touches its own UI. Effectively invisible.
Changes what the page sees. Blockers, style injectors, script managers. Findable, and worth knowing that ad blockers are extremely common on real browsers, so their presence is not by itself unusual. Their absence is more unusual than people assume.
Rewrites browser APIs. The fingerprinting-protection kind. These are the ones that create the arguing-with-the-engine problem.

The second point deserves emphasis, because it cuts against the instinct. A very large share of real browsers run an ad blocker. A profile that carries none is not suspicious on its own, but it is one more small way in which a fresh automated profile differs from a used personal one, alongside having no cookies.

How to load one, if you decide to

There is no extension argument in this wrapper. What there is is a persistent profile, and an extension installed into that profile stays there:

with InvisiblePlaywright(seed=42, profile_dir="/path/to/profile") as browser:
    ...

Launch once with that path, install the .xpi the normal way through the browser’s own add-ons page, then reuse the directory. Without profile_dir, every session builds a fresh profile from the seed, which is exactly why an extension installed in one run is gone in the next.

Pair a stable profile with a stable seed. A profile that accumulates cookies and storage while the hardware it claims changes every launch is a session whose history and machine disagree about how long it has existed.

Checking what yours is exposing

// does a known resource resolve?
fetch('moz-extension://<id>/manifest.json').then(r => console.log('present', r.status))
                                           .catch(() => console.log('not found'));

// did something remove my bait element?
const bait = document.createElement('div');
bait.className = 'ad-banner ads adsbox';
bait.style.height = '10px';
document.body.appendChild(bait);
setTimeout(() => console.log('blocked:', bait.offsetHeight === 0), 100);

Then run the same checks with no extensions loaded and compare. Anything that differs is what the extension is telling the page about you.

Short answers to the questions that lead here

Can websites detect browser extensions? Yes, through web-accessible resources, through what the extension does to the page, and through the traces any API override leaves.

Does using an ad blocker make me more detectable? It makes you findable, not unusual. Ad blockers are very common on real browsers.

Should I use a stealth extension with a stealth browser? No. Two layers deciding the same values contradict each other, and the contradiction is a stronger signal than either.

How do I add an extension in Playwright with Firefox? Through a persistent profile: install it once into that profile, then reuse the directory.

Why does my extension disappear between runs? Because a fresh profile is generated each launch unless you pass a persistent one.

Are extension IDs the same everywhere? In Firefox they are typically per-profile, which limits blind scanning. Do not rely on that as protection; extensions that need page access tend to expose predictable resources anyway.

See also: Function.prototype.toString and the native code check, which is what a page-level extension runs into, and three ways to make Playwright undetected, for why the layer an extension lives at has a ceiling.

From the notes of invisible_playwright, a Firefox patched at the C++ level. This page exists because someone asked in our discussions and got an answer about a different kind of extension entirely.