I’ve been running into this constantly. We’ll set up what feels like solid Playwright tests, and then a designer tweaks the layout or a backend change shifts how content loads, and everything falls apart. The selectors become invalid, timing breaks, and suddenly I’m spending hours rewriting test logic instead of actually testing.
I get that brittle tests are kind of a known problem, but I’m wondering if there’s something fundamentally different about how we should be writing these workflows. Like, are we supposed to be building in more resilience from the start, or is there a tool that actually helps with this?
What’s your approach when you hit this? Do you just accept that maintenance is going to be painful, or have you found a way to make tests adapt when things change?
This is where AI-driven workflow generation changes everything. Instead of hand-coding brittle selectors, you describe what the test should do in plain English, and the system generates Playwright workflows that adapt when the UI shifts.
The key difference is that AI-generated workflows can learn from context rather than just relying on fragile element selectors. When a layout changes, you regenerate the workflow description, and it pulls from the actual current state of the page.
I’ve seen teams cut maintenance time by like 60% because they’re not chasing selector changes anymore. They just update the description and let the AI handle the Playwright steps.
The real issue is that we’re thinking about this backwards. Most people build tests around what they see on screen right now, but sites are constantly evolving. I started using AI to generate the core workflow structure, which forces you to think about what the test is actually supposed to verify, not just which element to click.
Turns out when you describe a test in plain language instead of hand-coding selectors, the AI workflow is way more resilient to UI changes. It’s still Playwright under the hood, but the abstraction layer makes a huge difference.
I’ve dealt with this extensively. The fundamental problem is selector fragility, but there are patterns that help. What I found effective is building workflows that rely on semantic structure rather than pure CSS selectors. Instead of targeting specific divs, you target form labels, button text, and page structure. This means even when designers reorganize things, your tests still work because you’re testing the actual interface contract, not the implementation details.
use more generic selectors like aria-labels or data attributes instead of class names. they dont change as often. also consider ai workflow generation to handle reselection automaticly when uis update.