Automation Testing interview questions and answers

1. What HTTP/3 changes

HTTP/3 runs over QUIC instead of TCP. What is one key benefit?

1. A.A lost packet does not stall the other streams
2. B.It removes the need for HTTPS certificates
3. C.It makes JavaScript files run faster
4. D.It caches files forever in the browser

2. Sharing test builds with QA

QA and the client need new Android and iOS test builds several times a day without going through store review. Which tool fits?

1. A.Uploading the builds to production with a beta label
2. B.Emailing the source code to each tester
3. C.Firebase App Distribution or TestFlight internal testing
4. D.Posting the builds on a public website

3. Data-driven testing

You have one login test and a CSV file with 30 rows of username, password and expected message. Running the same steps once per row is called what?

1. A.Exploratory testing
2. B.Data-driven testing
3. C.Monkey testing
4. D.Smoke testing

4. Choosing a stable locator

Which locator is usually the most stable for clicking a "Submit" button in a UI test?

1. A.An absolute XPath like /html/body/div[2]/div/form/button[3]
2. B.A CSS class generated by the build like .btn_x7f2a
3. C.A role-based or test-id locator, for example getByRole('button', { name: 'Submit' }) or [data-testid="submit"]
4. D.The pixel position of the button on the screen

5. What a pytest fixture does

What is the main purpose of a pytest fixture such as the one below?

@pytest.fixture
def api_client():
    client = ApiClient(base_url=BASE_URL)
    yield client
    client.close()

1. A.It marks the test as skipped
2. B.It provides reusable setup to tests that ask for it by name, and runs the code after yield as teardown
3. C.It runs the test in parallel automatically
4. D.It turns the function into a test case

6. BeforeEach or BeforeAll

In JUnit 5, what is the difference between @BeforeEach and @BeforeAll?

1. A.@BeforeEach runs once per test class; @BeforeAll runs before every test
2. B.They are the same annotation with different names
3. C.@BeforeEach runs before every test method; @BeforeAll runs once before all tests in the class
4. D.@BeforeAll only runs when a test fails

7. How Selenium talks to a browser

How does Selenium WebDriver control a browser?

1. A.It sends commands through the W3C WebDriver protocol to a browser driver such as chromedriver or geckodriver
2. B.It records the screen and replays mouse movements
3. C.It edits the browser source code at runtime
4. D.It only works by injecting jQuery into the page

8. Screenshots on failure

Why configure the framework to save a screenshot when a UI test fails?

1. A.Screenshots make the test pass on retry
2. B.It is required by WCAG
3. C.It shows what the page looked like at the moment of failure, which speeds up debugging
4. D.It lowers the cost of the CI server

9. Arrange, Act, Assert

Which description matches the Arrange-Act-Assert pattern for structuring a test?

1. A.Set up data and objects, run the one behaviour under test, then check the outcome
2. B.Write the assertion first, then the code, then refactor the test
3. C.Run every action in the system, then assert all results at the end
4. D.Arrange tests by priority, act on the failures, then assign owners

10. A double that returns fixed data

In a unit test, an exchange-rate service is replaced by an object that always returns 22.85 for AED to INR, with no checks on how it was called. What kind of test double is this?

1. A.A mock
2. B.A stub
3. C.A spy
4. D.A dummy

11. Reading a Gherkin scenario

In a Gherkin scenario, what does each keyword describe?

Scenario: Free delivery for large orders
  Given a cart worth ₹2,500
  When the customer checks out
  Then delivery should be free

1. A.Given is the starting context, When is the action, Then is the expected outcome
2. B.Given is the expected outcome, When is the setup, Then is the action
3. C.Given is the test name, When is the tag, Then is the data table
4. D.Given lists the browsers, When lists the pages, Then lists the locators

12. A failure message that helps

A CI log shows only "AssertionError: expected false to be true". What change would make such failures faster to diagnose?

1. A.Assert on the actual values, e.g. expect(status).toBe("PAID"), or add a clear message
2. B.Wrap every assertion in try/catch and print "test failed" in the catch block
3. C.Turn the assertion into a console.log so that the run never stops there
4. D.Run the failing test again in CI with retries until it passes

13. Expecting an exception in JUnit 5

In JUnit 5, how do you check that withdraw(500) on an account with ₹100 throws InsufficientFundsException and then inspect its message?

1. A.@Test(expected = InsufficientFundsException.class) on the test method
2. B.var ex = assertThrows(InsufficientFundsException.class, () -> account.withdraw(500));
3. C.assertEquals(InsufficientFundsException.class, account.withdraw(500));
4. D.assertTrue(account.withdraw(500) instanceof InsufficientFundsException);

14. Finding a button by role and name

A page has several buttons. Which Playwright locator best finds the one labelled "Place order" in a way that also reflects how users and screen readers see it?

1. A.page.getByRole('button', { name: 'Place order' })
2. B.page.locator('div > div:nth-child(3) > button')
3. C.page.locator('//html/body/div[2]/form/button[1]')
4. D.page.locator('.btn.btn-primary.mt-2')

15. Uploading a file

How do you upload a file through an <input type="file"> in Playwright?

1. A.await page.getByLabel('Invoice').fill('fixtures/invoice.pdf')
2. B.await page.keyboard.type('C:\\fixtures\\invoice.pdf')
3. C.await page.getByLabel('Invoice').setInputFiles('fixtures/invoice.pdf')
4. D.await page.getByLabel('Invoice').click({ file: 'invoice.pdf' })

16. A menu that opens on hover

A sub-menu item only appears when the mouse hovers over "Reports". How do you click it in Playwright?

1. A.Click the sub-menu item directly with { force: true }
2. B.Change the CSS in the test so that the sub-menu is always visible
3. C.Wait five seconds so that the sub-menu has time to appear on its own
4. D.Hover over 'Reports' first, then click the sub-menu item's locator

17. Keeping videos only for failures

You want videos for failed tests but not for passing ones, to save CI storage. Which setting fits?

1. A.video: 'retain-on-failure'
2. B.video: 'on-first-retry'
3. C.video: 'off'
4. D.screenshot: 'only-on-failure'

18. The Select helper in Selenium

Which Selenium approach chooses "Abu Dhabi" from a native <select id="emirate">?

1. A.driver.findElement(By.id("emirate")).sendKeys(Keys.ENTER, "Abu Dhabi")
2. B.driver.findElement(By.id("emirate")).setValue("Abu Dhabi")
3. C.new Select(driver.findElement(By.id("emirate"))).selectByVisibleText("Abu Dhabi")
4. D.driver.switchTo().select("emirate").choose("Abu Dhabi")

19. What Selenium Grid is for

What is Selenium Grid mainly used for?

1. A.Running tests on remote machines and browsers, often in parallel
2. B.Recording tests by watching a manual tester's clicks
3. C.Comparing screenshots against stored baseline images
4. D.Generating test data for forms automatically

20. Reading ease scores

A readability tool shows a Flesch Reading Ease score. What does a higher score mean?

1. A.The text is longer
2. B.The text has more keywords
3. C.The text is harder to read
4. D.The text is easier to read

21. toBe fails on equal objects

Why does this test fail even though both objects look the same?

expect(getUser()).toBe({ id: 1, name: 'Ana' })

1. A.toBe only works with strings
2. B.toBe uses Object.is, which compares object references, so a different object never matches; use toEqual
3. C.Vitest cannot compare objects that contain strings
4. D.The test must be marked async to compare objects

22. Flaky end-to-end test

This Playwright test passes locally but fails randomly in CI. What is the best fix?

await page.getByRole('button', { name: 'Save' }).click()
await page.waitForTimeout(2000)
expect(await page.getByText('Saved').isVisible()).toBe(true)

1. A.Increase the timeout to 10000 ms
2. B.Use await expect(page.getByText('Saved')).toBeVisible(), which auto-retries
3. C.Run the tests with retries set to 5
4. D.Mock the whole backend with a static JSON file so the timing is always identical

23. Flaky UI test timing

An Espresso or XCUITest test fails randomly because it checks a label before a network-backed screen has finished loading. What is the best fix?

1. A.Add a fixed sleep of a few seconds before every check, tuned on the slowest CI machine, so the screen always has time to finish loading first
2. B.Wait for a condition (idling resources in Espresso, waitForExistence in XCUITest, waitUntil in Compose tests) and replace the real network with a fake
3. C.Retry the test until it passes and ignore failures
4. D.Delete the test, since UI tests are always flaky

24. Native module undefined in Jest

A Jest test fails with "NativeModule: AsyncStorage is null" when rendering a screen. What is the standard fix?

1. A.Run Jest on a real Android device instead of in Node
2. B.Install the native module again with pod install
3. C.Mock the native module, for example with the library's mock
4. D.Switch the test environment to jsdom for the web

25. Same test, many inputs

You want to run the same login validation test with 12 different invalid inputs. What is the cleanest approach?

1. A.Copy the test 12 times with small edits
2. B.Put all 12 inputs in one test and stop at the first failure
3. C.Test only the first input
4. D.Use a parameterised (data-driven) test, for example pytest.mark.parametrize or a loop that generates one test per row

26. Click intercepted by an overlay

A test fails with "element click intercepted: another element would receive the click". A loading overlay fades out after data loads. What is the best fix?

1. A.Use force: true so the click ignores the overlay
2. B.Add a 3 second sleep before the click
3. C.Wait for the overlay to be hidden (or for the loaded data to appear), then click normally
4. D.Click with JavaScript using element.click() in the browser

27. Splitting a suite across machines

What does npx playwright test --shard=2/4 do?

1. A.Runs each test 2 times on 4 browsers
2. B.Uses 2 workers and 4 retries
3. C.Runs the second quarter of the test files, so 4 machines together run the whole suite
4. D.Runs only tests whose names contain 2 or 4

28. Passes in the suite, fails alone

A test passes when the whole suite runs but fails when run on its own. What does this most likely show?

1. A.The test runner is broken
2. B.The test is too short
3. C.The test should be deleted
4. D.The test depends on state created by an earlier test, so it is not isolated

29. Tests for several user roles

You need tests for admin and viewer users. Logging in through the UI in every test is slow. What is a clean approach in Playwright?

1. A.Share one browser tab between all tests
2. B.Run a setup project that logs in each role once and saves separate storage state files; tests choose the file for their role
3. C.Turn off authentication in production
4. D.Log in as admin for all tests and hide admin controls with CSS

30. A test that passes for the wrong reason

This test passes even when fetchUser rejects. Why, and what is a fix?

test('loads user', () => {
  fetchUser(7).then((user) => {
    expect(user.name).toBe('Asha')
  })
})

1. A.Jest ignores assertions inside then(), so move it to an afterEach hook
2. B.toBe cannot compare strings, so it should use toEqual instead
3. C.The promise is not returned or awaited; use async and await fetchUser(7)
4. D.The test needs a longer timeout so the promise can finish

31. The app uses data-qa, not data-testid

Your app marks elements with data-qa="..." instead of data-testid. How do you make page.getByTestId() use it?

1. A.Rename every attribute in the app to data-testid before testing
2. B.Use getByTestId('[data-qa=submit]') with the full attribute selector
3. C.getByTestId cannot be changed, so use XPath for these elements
4. D.Set use: { testIdAttribute: 'data-qa' } in playwright.config

32. Changing a request on its way out

You want real API calls to go through, but with an extra header x-feature: beta. Which route handler does that?

await page.route('**/api/**', async (route) => {
  // ?
})

1. A.route.continue({ headers: { ...route.request().headers(), 'x-feature': 'beta' } })
2. B.route.fulfill({ status: 200, headers: { 'x-feature': 'beta' }, body: '{}' })
3. C.route.abort('failed'); route.continue({ headers: { 'x-feature': 'beta' } })
4. D.route.fulfill({ headers: { ...route.request().headers(), 'x-feature': 'beta' } })

33. Picking one product card

A product grid has many cards with an "Add to cart" button each. Which locator clicks the button inside the card for "Basmati Rice 5kg"?

1. A.page.getByTestId('product-card').first().getByRole('button', { name: 'Add to cart' })
2. B.page.getByText('Basmati Rice 5kg').getByRole('button', { name: 'Add to cart' })
3. C.page.getByRole('button', { name: 'Add to cart' }).nth(4).filter({ visible: true })
4. D.page.getByTestId('product-card').filter({ hasText: 'Basmati Rice 5kg' }).getByRole('button')

34. First run of a screenshot test

A new test uses await expect(page).toHaveScreenshot(). On its very first run, with no baseline image, what happens?

1. A.It passes and silently compares the page against a blank image
2. B.It downloads a matching baseline image from the Playwright website
3. C.It skips the assertion until someone uploads a reference image
4. D.It fails, saves the actual image as the baseline, and passes next time

35. Checking the number of results

After applying a filter, the list should show exactly 3 items, but it updates asynchronously. Which check is most reliable?

1. A.await expect(page.getByRole('listitem')).toHaveCount(3)
2. B.expect(await page.getByRole('listitem').count()).toBe(3)
3. C.expect((await page.getByRole('listitem').all()).length).toBe(3)
4. D.expect(await page.locator('li').isVisible()).toBe(true)

36. A wait that ignores early errors

A status label appears after a few seconds. Which Selenium FluentWait setup waits up to 20 seconds, checks every 500 ms and ignores "not found" errors while waiting?

Wait<WebDriver> wait = new FluentWait<>(driver)
    .withTimeout(Duration.ofSeconds(20))
    .pollingEvery(Duration.ofMillis(500))
    .ignoring(NoSuchElementException.class);

1. A.The same code but without ignoring(), as FluentWait ignores all errors anyway
2. B.Thread.sleep(20000) followed by one findElement call
3. C.This one: timeout, polling interval and ignored exception set as shown
4. D.driver.manage().timeouts().pageLoadTimeout(Duration.ofSeconds(20))

37. The risk of @CacheLookup

In a Selenium PageFactory page object, adding @CacheLookup to a field made a test fail with StaleElementReferenceException after a list refreshed. Why?

1. A.The element was found once and reused, so the replaced DOM node went stale
2. B.@CacheLookup makes Selenium look up the element on every single call
3. C.@CacheLookup only works with XPath locators, not with CSS selectors
4. D.PageFactory fields cannot be used once the page has been reloaded

38. A date picker used on ten pages

The same date picker widget appears on ten different pages. What is a good way to model it in your page objects?

1. A.Copy the date picker methods into each of the ten page objects
2. B.A small component object, created with a root locator, reused by each page object
3. C.Put all ten pages into one huge page object so the methods are shared
4. D.Avoid modelling it and write raw locators in each test instead

39. Checking a sorted column

After clicking the "Amount" header, the table should sort by amount ascending. Values look like "₹1,200.00". What is a sound way to verify it?

1. A.Check that the first row has the smallest amount and ignore the rest
2. B.Compare the cell strings with a copy sorted alphabetically as text
3. C.Check that a sort arrow icon is shown on the Amount header
4. D.Read all amount cells, parse them to numbers, and compare with a sorted copy

40. Organic landing pages in GA4

In GA4, how do you see which landing pages bring the most organic search sessions and conversions?

1. A.Use the Realtime report only
2. B.Open the Landing page report and filter or compare by session default channel group equal to Organic Search
3. C.Look at the Demographics report
4. D.GA4 cannot show organic data; only Search Console can

41. Noisy screenshot tests

Visual regression tests fail on most runs with tiny pixel differences in fonts, plus a changing date and a carousel. Developers now ignore the failures. What is the best fix?

1. A.Raise the allowed pixel difference to 30% for all tests
2. B.Delete all screenshot tests
3. C.Run screenshots in one fixed environment (same container, browser and fonts), mask or freeze dynamic content, and disable animations
4. D.Take screenshots only on developers' own laptops

42. When automation pays off

A regression set takes 4 hours to run by hand. Automating it takes 40 hours, and each automated run needs 1 hour of upkeep and analysis. After how many runs does automation start saving time?

1. A.10 runs
2. B.40 runs
3. C.14 runs
4. D.8 runs

43. Element inside an iframe

In Selenium, a payment card field is inside an iframe, and findElement(By.id("card")) throws NoSuchElementException. What is needed?

1. A.Switch to the frame first with driver.switchTo().frame(...), interact, then switch back with defaultContent()
2. B.Use a longer implicit wait
3. C.Use an absolute XPath from the top document
4. D.Run the test in headless mode

44. Waiting for network idle

A team waits for "networkidle" after every navigation. Tests are slow and still flaky on pages with analytics and polling. What is the better approach?

1. A.Increase the networkidle timeout to 60 seconds
2. B.Block all network requests
3. C.Wait for the specific UI state the test needs, using web-first assertions on visible content
4. D.Use page.waitForTimeout(5000) instead

45. DRY or DAMP tests

A framework has helpers so abstract that a test reads: runScenario(config[7]). Reviewers cannot tell what it checks. What guideline applies?

1. A.Tests should be as DRY as possible, even if unreadable
2. B.Prefer DAMP (descriptive and meaningful phrases) in tests: some repetition is fine if each test clearly shows its setup, action and expected result
3. C.Remove all helpers and copy code everywhere
4. D.Move all assertions into the helpers

46. Testing with generated inputs

A function splits a bill among N people in whole paise. Instead of a few hand-picked examples, you want to check that the shares always add up to the total for many random inputs. Which approach is this?

1. A.Snapshot testing of the output for one large input
2. B.Mutation testing of the function's source code
3. C.Contract testing between the bill service and its users
4. D.Property-based testing, for example Hypothesis or fast-check

47. Shuffling test order on purpose

Why might a team run tests in random order (for example with pytest-randomly or jest --randomize) in CI?

1. A.To make the suite run faster by spreading slow tests evenly
2. B.To expose hidden dependencies between tests, with a seed to replay
3. C.To give every test in the suite an equal chance of being skipped
4. D.To make results differ each time so flaky tests are hidden

48. SQLite in tests, Postgres in production

Tests run against SQLite for speed while production uses PostgreSQL. Which bug is most likely to slip through?

1. A.A type, constraint or collation rule that SQLite handles differently
2. B.A typo in a local variable name inside the repository class
3. C.A missing import of the database driver in the main application code
4. D.A wrong HTTP status code returned by the API controller

49. A cache that remembers old tests

A singleton PriceCache is filled by one test with a ₹999 price. Another test later expects ₹1,099 from a stubbed service but gets ₹999. What is the best long-term fix?

1. A.Order the tests so that the ₹1,099 test always runs first
2. B.Inject the cache so each test gets a fresh one, or reset it in a fixture
3. C.Add a sleep so the cache entry expires before the second test
4. D.Mark the second test as flaky and allow two retries

50. A test that can never fail

Why does this test always pass, even when no "Overdue" invoice exists?

test('flags overdue invoices', () => {
  const invoices = getInvoices()
  invoices.filter((i) => i.status === 'Overdue').forEach((i) => {
    expect(i.reminderSent).toBe(true)
  })
})

1. A.If the filter returns no items, no assertion runs and the test passes
2. B.forEach runs asynchronously, so Jest misses the assertions inside
3. C.toBe(true) always passes when the property is undefined
4. D.Jest skips tests that use filter() on arrays of objects

51. Checking structure, not pixels

Screenshot tests of a navigation menu break with every font or colour change, but the team mainly cares that the right links exist in the right order. Which check fits better?

1. A.toHaveScreenshot() with a 50% pixel difference tolerance
2. B.Asserting that the menu's innerHTML equals a saved string
3. C.expect(locator).toMatchAriaSnapshot() comparing the accessibility tree
4. D.Comparing the CSS class list of every link to a saved list

52. Paying one invoice in a table

In a table of invoices, each row has one button, "Pay", and rows = page.getByRole("row"). Which locator finds "Pay" only in the row whose invoice cell is exactly "INV-1043" (and not "INV-10431")?

1. A.rows.filter({ has: page.getByRole('cell', { name: 'INV-1043', exact: true }) }).getByRole('button')
2. B.rows.filter({ hasText: 'INV-1043' }).first().getByRole('button', { name: 'Pay', exact: true })
3. C.page.getByText('INV-1043').locator('..').getByRole('button', { name: 'Pay' }).first()
4. D.rows.getByRole('button', { name: 'Pay', exact: true }).nth(1043).filter({ hasText: 'Pay' })

53. Extra words in a table cell

A sortable column header shows "Amount", but allTextContents() returns "Amount sorted ascending" because of a <span style="display:none">. Which method returns only the rendered text?

1. A.allTextContents()
2. B.allInnerTexts()
3. C.getAttribute('textContent')
4. D.evaluate((el) => el.outerHTML)

54. Clicking through JavaScript

A tester "fixes" an intercepted click by using ((JavascriptExecutor) driver).executeScript("arguments[0].click()", button). Why can this be harmful?

1. A.It clicks even when a real user could not, so it can hide a real bug
2. B.JavaScript clicks are always slower than WebDriver clicks
3. C.executeScript does not work on buttons, only on links
4. D.It permanently disables the overlay for all other tests

55. Logging in on another domain in Cypress

The app redirects to an identity provider at login.idp.com for SSO. In Cypress, commands on that page fail. What is the supported way to interact with it?

1. A.Set chromeWebSecurity to false and use cy.visit for every page
2. B.Open the identity provider in a new tab with cy.window().open()
3. C.Wrap those steps in cy.origin('https://login.idp.com', () => { ... })
4. D.Use cy.wait(5000) so Cypress can switch to the other domain

56. Finding an item in an infinite feed

A feed loads 20 more items each time you scroll to the bottom. A test must find a post created a week ago. Which approach is most robust?

1. A.Scroll with an unbounded while loop until the post finally appears
2. B.Scroll once and wait 30 seconds for all the posts to load
3. C.Create the post in the test, or scroll in a bounded loop until found
4. D.Zoom the browser to 10% so every post fits on the screen

57. getText() returns an empty string

A hidden <span> with display:none holds a tracking number. Selenium's getText() returns "". What explains this, and how do you read it?

1. A.The element is stale; find it again and call getText() once more
2. B.getText() fails on spans; use getTagName() to read the value
3. C.Hidden text is encrypted; use executeScript to decrypt it
4. D.getText() returns only visible text; read getDomProperty("textContent")

58. Article text on a phone

On a 360 px wide phone, an article uses 13 px text with 8 px side margins to fit more words per line. What is the main problem and fix?

1. A.Lines are too short; use 11 px text
2. B.Margins are too wide; remove them
3. C.Line height is too large; set it to 1.0
4. D.Text is too small; use 16 px+ and wider margins

59. Setting text for readers with dyslexia

A government portal wants body text that is easier for readers with dyslexia. Which change has the strongest support from guidance such as the British Dyslexia Association style guide?

1. A.A special dyslexia font set in all caps
2. B.Fully justified text in narrow columns
3. C.Light grey text on pure white for softness
4. D.Left-aligned, well-spaced text, no long italics

60. A sales cut-off problem

The year ends on 31 March. Goods were dispatched on 3 April, but the invoice was dated 30 March and recorded as a sale for the year. Control passes on dispatch. What is the effect, and which assertion is affected?

1. A.Revenue understated; completeness
2. B.No effect, because the invoice date decides the period
3. C.Revenue and receivables overstated (and inventory understated); cut-off
4. D.Expenses overstated; classification