Headless runtime mismatches
A few page details look odd when a browser has no real window or desktop.
Older headless Chrome showed an empty plugin list, no languages, a zero outer window size and a notification setting that disagreed with the Permissions API. New headless mode closed most of these gaps. A software graphics renderer and a default 800 by 600 window can still stand out.
What it can and can't do
What the user does: Nothing. Page script runs when the page loads
It can
- Compare window size, plugins and languages with what a desktop shows
- Compare two permission answers that should match
- Notice a software graphics renderer on a server
- Raise a bot score from several small mismatches
It cannot
- Prove automation from any one value
- Catch new headless mode on its own
- See what the agent is doing
- Resist a stealth plugin that patches each value
How it differs between platforms: Newer Chrome builds fixed several of these. A stealth plugin can patch the rest, so sites treat them as weak clues and combine them.
Browser support
Version notes
- Chrome: Applies to headless Chrome
- Edge: Applies to headless Edge
- Firefox: No known headless specific values
- Safari: No headless mode
How a site asks for it
| API | navigator.plugins.length; navigator.languages; window.outerWidth; Notification.permission vs permissions.query({name:'notifications'}); WebGL renderer; window.chrome |
| Browser compat keys | api.Navigator.pluginsapi.Navigator.languagesapi.Window.outerWidthapi.Permissions.query |
| How access is granted | none |
| What the user sees | None |
| Needs a secure context (HTTPS) | No |
| Needs a user gesture | No |
| How long it lasts | Not listed |
| How to take it back | Not listed |
| Inside an iframe | Not listed |
Typical flow
- Page script reads window, navigator and permission values
- Script cross checks them against each other
- Mismatches add to a bot score
Data you get
Shape: object
How often: static per session
| Field | Type | Unit | Range | Precision |
|---|---|---|---|---|
plugins.length | integer | None | 0 in older headless Chrome | Not listed |
languages | list | None | empty in older headless Chrome | Not listed |
outerWidth / outerHeight | integer | px | 0 in older headless Chrome | Not listed |
Notification.permission vs permissions.query | enum | None | denied vs prompt mismatch in older headless Chrome | Not listed |
WebGL renderer | string | None | software renderer such as SwiftShader on servers | Not listed |
Technical limits
- Several values changed in newer headless mode
- Patched by stealth plugins
- A real server hosted desktop browser can show some of them
The story
What it is
A set of tiny oddities that a screenless browser leaves behind when its values do not agree with each other.
For you
You never see it. A normal browser agrees with itself everywhere, so nothing happens. A headless agent can fail a few small tests.
For makers
Treat each value as a weak hint. Contradictions between values (a permission that says denied in one API and prompt in another) are stronger than any single value. Re check after Chrome releases, because defaults change.
Where you see it
| Type | Example | Apps |
|---|---|---|
| security | A bot filter compares two permission answers and flags a mismatch. | Cloudflare |
| shopping | A sneaker shop challenges clients with no plugins and no languages. | None named |
| ads | An ad verifier drops views from clients with a software graphics renderer on a datacenter. | None named |
| games | A browser game checks window size to catch bots that run without a screen. | None named |
| analytics | Analytics filters visits with a zero sized window. | None named |
| developer | A tester runs a detection page to see how their headless setup looks. | Puppeteer, Playwright |
Risks
- False positives for people on unusual setups
- Easy to patch
What the platform says
- Chrome for Developers: Headless Chrome. Docs describe Chrome running without a visible UI; no page window exists.
Good uses and abuse
Legitimate uses
- Bot defense
- Fraud scoring
Known abuse
- Blocking legitimate agents
- Penalizing people on remote desktops
Scores
| Score | Value | Meaning |
|---|---|---|
| Sensitivity | 1 of 5 | How revealing the data is |
| Utility | 3 of 5 | How useful it is to honest apps |
| Friction | 0 of 5 | How much the user has to do |
| Fingerprint bits | 1 | Rough identifying entropy, when known |
History
- 2017-06, Chrome 59: Headless mode ships with several values that differ from headed Chrome
- 2023, Chrome 112: New headless mode brings it much closer to headed Chrome
Related capabilities
Related
- Headless label in the browser identity. Can show that Chrome is running without a visible window
- Automation (webdriver) flag. Can tell if Selenium, Playwright, or Puppeteer is controlling the browser
- navigator.plugins and mimeTypes. Can see a fixed list of 5 PDF viewers, so it no longer tells you apart
- GPU via WebGL unmasked renderer. Can see your graphics card and driver name through WebGL
- Permissions API. Can check whether you allowed or blocked a feature, without asking you
Sources
How well checked: Inferred from related behavior.