browser automation testing tools
Four names come up constantly and get compared as if they were interchangeable: Playwright, Selenium, Cypress and Puppeteer. They overlap enough to confuse and differ enough that picking by feature-list length is how people end up fighting the wrong tool for a year.
Playwright. General-purpose, cross-browser (Chromium, Firefox, WebKit), built by Microsoft with auto-waiting and multi-language support (JS/TS, Python, Java, .NET). The current default recommendation for new browser-automation projects that need real cross-browser coverage.
Selenium. The oldest and most widely deployed, via the W3C WebDriver standard. Broadest language support, largest historical ecosystem, and the one that can drive an actual retail browser install rather than a framework-maintained build - relevant if you specifically need to test against exactly what your users have installed.
Cypress. Built specifically for testing web applications from inside the browser, running in the same run-loop as your app rather than driving it from outside. Excellent developer experience and debugging for single-page apps, historically Chromium-only in practice (with newer versions adding limited Firefox and WebKit support), and architecturally not built for the kind of cross-origin, multi-tab scraping work Playwright and Selenium both handle natively.
Puppeteer. Chromium/Chrome only, Node.js only, maintained by the Chrome team. If your target is specifically Chrome and your stack is Node, it is a direct, well-supported option; it does not cover Firefox or WebKit at all.
| Your situation | Reach for |
|---|---|
| Cross-browser regression suite, multiple languages on the team | Playwright |
| Testing a Chrome-only internal app, Node stack already in place | Puppeteer |
| Fast-feedback testing of a single-page app you are actively developing | Cypress |
| Must drive an actual retail-installed browser, not a framework build | Selenium |
| Large legacy Selenium suite that works | Selenium, don't rewrite for its own sake |
| Real-world scraping across multiple sites, need Firefox specifically | Playwright |
All four of these are testing tools, built to verify that your own application behaves correctly under known, cooperative conditions. None of them was designed against a page actively trying to distinguish automation from a human - because for their primary use case, the page being tested is yours, and it has no reason to fight you.
That changes completely when the target is someone else's site and it is
actively fingerprinting. Every one of these four, used with its default
browser build, presents recognizable automation signals: navigator.webdriver
under WebDriver-based tools, CDP-visible properties under CDP-driven ones,
and browser builds that differ from a retail install in ways a determined
detector checks for. Picking Playwright over Selenium, or Puppeteer over
Cypress, does not change this axis at all - it is orthogonal to which testing
tool you picked.
invisible_playwright is not a fifth entry on the list above competing for the same job. It answers a different question: given that you have already picked Playwright's API (unchanged, same code you would write against any Playwright browser), what if the Firefox underneath it was patched at the C++ source rather than Playwright's stock automation build? It is the engine layer beneath a testing tool, not a replacement for one. Firefox or Chromium for anti-detect automation covers why Firefox was the base worth patching, and patchright vs invisible_playwright covers the equivalent idea applied to Chromium instead.
Is Playwright better than Selenium? For new cross-browser projects, usually yes - faster, auto-waiting, fewer moving parts. Selenium's breadth of language support and ability to drive a real retail browser remain real advantages in specific cases.
Is Cypress a replacement for Selenium or Playwright? Not a direct one. Cypress is purpose-built for testing an application you control from inside the browser's run-loop, and is a weaker fit for cross-origin scraping or multi-tab flows that Selenium and Playwright handle natively.
Does Puppeteer support Firefox? Puppeteer has had experimental Firefox support at various points; it is not the primary, well-maintained target the way Chrome is. If Firefox specifically matters, Playwright is the better-supported cross-browser option.
Which tool is hardest to detect? None of the four is built with detection resistance as a design goal, and the default automation build under any of them is recognizable. That is a separate axis from which testing tool fits your testing needs.
Can I use these for scraping instead of testing? Yes, and Playwright and Selenium in particular are commonly used this way - but recognize that you have moved outside each tool's primary design intent, onto an axis (adversarial detection) none of them was built to address.
See also: Selenium vs Playwright: the actual differences, Migrating from Puppeteer to Playwright for stealth, and Firefox or Chromium for anti-detect automation.
- Playwright's own documentation, Selenium's documentation, Cypress's documentation, and Puppeteer's documentation, all retrieved 2026-09-10, for each project's stated scope and browser support.
Written while maintaining invisible_playwright, which is not a competitor to any of the four tools above - it sits underneath one of them. The page is organized around picking the right testing tool first, because that decision is independent of the engine question it exists to answer.