Conversational Test Creation: Prompting is the new Page Object Model

For over a decade, the Page Object Model (POM) has been the undisputed king of UI automation design patterns. By abstracting web elements into classes, QA engineers made tests readable and maintainable. But in 2026, a new abstraction layer has arrived that makes POM look like assembly language: Natural Language Prompting.

The End of Manual Locators

In a conversational testing framework, you don't inspect the DOM. You don't write `page.locator('#submit-btn')`. Instead, you interact with an agentic testing tool (like Auto Playwright or a custom MCP Server) using plain English.

// ❌ 2023: The Page Object Model Approach
const loginPage = new LoginPage(page);
await loginPage.enterUsername('admin');
await loginPage.enterPassword('secret123');
await loginPage.clickSubmit();
await expect(page.locator('.welcome-msg')).toBeVisible();

// ✅ 2026: The Conversational Approach
await auto(page, 'Log in using admin and secret123, then verify the welcome message appears.');

How it Works Under the Hood

The "Prompt Engineer" QA Role

If tests are written in English, does that mean anyone can write them? Yes and no. While product managers can now script happy paths easily, QA Engineers are shifting into a "Prompt Engineering" role.

Writing "Test the checkout" is too vague for an LLM to reliably execute edge cases. A modern QA writes prompts like: *"Attempt to submit the checkout form with an invalid email format. Verify the specific inline validation error 'Please enter a valid email' appears in red beneath the input."* Precision in natural language is the new precision in CSS selectors.

Conclusion

The Page Object Model isn't entirely dead for highly secure, complex calculations. But for 80% of standard UI flows, conversational test creation is faster to write, completely resilient to DOM changes, and self-maintaining. The future of QA is chatting with the browser.