Playwright's built-in locator tools like getByText and getByRole often struggle with modern component libraries. Many applications use custom UI elements built with divs and hidden inputs instead of semantic HTML. This makes standard assertions for checkboxes or disabled states unreliable, as the underlying DOM doesn't expose the expected attributes.

When standard locators fail, developers must craft their own. The most reliable strategy is anchoring to stable text labels and navigating the DOM. Playwright's locator('..') API allows walking up and down the hierarchy cleanly. This approach is more precise than XPath and lets you target the actual elements representing state, like aria-checked attributes.

For example, locating a custom checkbox involves chaining locators from a text anchor up to a container and back down to the specific input. Asserting on concrete attributes like aria-checked and disabled proves more robust than relying on higher-level helpers. This method ensures tests mirror what users see, not just semantic HTML structures.

This practical limitation challenges claims that AI testing tools can fully automate test creation. Real-world apps are full of framework-specific markup and hidden inputs. For now, a developer who can anchor on text, traverse the DOM, and assert on real attributes remains essential for writing strong, reliable Playwright tests.