1. Accessibility Testing
Accessibility means people with disabilities (screen readers, keyboard only, low vision) can use your site. Automated tools find roughly a third of the problems very quickly, such as missing alt text, low contrast and missing form labels.
bash
pip install axe-selenium-pytho
python
from axe_selenium_python import Axe
def test_home_page_has_no_critical_accessibility_issues(driver):
driver.get("https://the-internet.herokuapp.com/")
axe = Axe(driver)
axe.inject() # adds the axe-core script to the page
results = axe.run()
serious = [v for v in results["violations"] if v["impact"] in ("critical", "serious")]
for v in serious:
print(v["id"], "-", v["help"], "-", len(v["nodes"]), "elements")
assert not serious, f"{len(serious)} serious accessibility problems found"
| Common finding | Fix |
Images without alt | Add a description |
| Form field without label | Add <label for> or aria-label |
| Low colour contrast | Use darker text or lighter background |
| Missing page language | <html lang="en"> |
| Buttons without text | Add visible text or aria-label |
Not a replacement for manual checks: Automatic scans cannot judge whether the page makes sense with a screen reader. Add a manual keyboard-only test (Tab, Enter, Escape) in your regular testing.
Keyboard Navigation Test
python
from selenium.webdriver.common.keys import Keys
def test_can_reach_login_button_with_tab(driver):
driver.get("https://www.saucedemo.com")
body = driver.find_element(By.TAG_NAME, "body")
for _ in range(3):
body.send_keys(Keys.TAB)
assert driver.switch_to.active_element.get_attribute("id") == "login-button"
2. Visual Regression Testing
Visual tests compare a new screenshot with an approved baseline. They catch layout problems that normal assertions miss (overlapping text, broken CSS).
bash
pip install pillow
python
from pathlib import Path
from PIL import Image, ImageChops
BASELINE = Path("visual/baseline")
ACTUAL = Path("visual/actual")
def assert_matches_baseline(driver, name, tolerance=0):
ACTUAL.mkdir(parents=True, exist_ok=True)
BASELINE.mkdir(parents=True, exist_ok=True)
actual_path = ACTUAL / f"{name}.png"
driver.save_screenshot(str(actual_path))
baseline_path = BASELINE / f"{name}.png"
if not baseline_path.exists():
actual_path.replace(baseline_path) # first run creates the baseline
return
a = Image.open(actual_path).convert("RGB")
b = Image.open(baseline_path).convert("RGB")
assert a.size == b.size, f"Image size changed: {a.size} vs {b.size}"
diff = ImageChops.difference(a, b)
bbox = diff.getbbox()
if bbox is not None:
changed = sum(1 for px in diff.getdata() if px != (0, 0, 0))
assert changed <= tolerance, f"{changed} pixels differ. See {actual_path}"
Make screenshots stable: Fix the window size, hide dates, ads and animations (inject CSS), wait for fonts and images, and use the same browser and OS as the baseline. Tools such as Applitools, Percy and BackstopJS handle this for you.
3. Performance Checks
python
def test_home_page_loads_fast(driver):
driver.get("https://www.selenium.dev")
nav = driver.execute_script("return performance.getEntriesByType('navigation')[0].toJSON()")
dom_ready = nav["domContentLoadedEventEnd"]
load_time = nav["loadEventEnd"]
print(f"DOM ready: {dom_ready:.0f} ms, full load: {load_time:.0f} ms")
assert load_time < 5000, f"Page took {load_time:.0f} ms to load"
For deeper analysis run Lighthouse (in Chrome DevTools or npx lighthouse <url>) in your CI pipeline and fail the build when the score falls under a limit.
Performance of the test run is not load testing: Selenium measures what one user sees. To simulate thousands of users use tools such as JMeter, k6 or Locust.
Conclusion
Add small automatic checks to your suite: axe for accessibility, screenshot comparison for layout and navigation timing for speed. They find problems that normal functional tests miss.
Quick Summary: axe = accessibility, image diff = visual, performance.timing = speed.