December 1, 2023 ADC Team 6 min read

AI and Machine Learning for Automated Testing

AI does not replace QA Engineers — it amplifies their impact. From auto-healing tools to LLMs capable of generating tests from specifications, AI is transforming every stage of the test cycle. Here is the state of the art in 2024.

Key takeaways

Auto-healing (Testim, Mabl): ML automatically detects renamed selectors and updates tests — no human intervention required for DOM changes
Applitools Eyes: AI-powered visual regression that understands page structure, ignoring anti-aliasing noise while catching genuine layout regressions
LLMs (GPT-4, Claude, Gemini) generate Playwright/Cypress test drafts from natural language — estimated 40–60% time saving on initial test creation
Intelligent test prioritisation (Launchable, Saucelabs): run 20% of tests, detect 80% of bugs by predicting which tests are most likely to fail for a given change

1. Auto-Healing — Tests That Fix Themselves

Auto-healing is the most immediately useful AI feature. When a selector changes (renamed ID, modified CSS class), the AI analyses the DOM and automatically finds the new corresponding element. Tools like Testim and Mabl use ML to keep tests up to date without human intervention.

2. AI Visual Analysis — Applitools

Applitools Eyes uses AI to visually compare pages at each run. Unlike simple screenshots, the AI understands the page structure and ignores minor differences (anti-aliasing, slightly different rendering) while detecting genuine visual regressions.

3. Test Generation with LLMs

LLMs (GPT-4, Claude, Gemini) can generate Playwright or Cypress tests from natural language descriptions. The quality is sufficient for a first draft — a QA Engineer reviews and refines. Estimated time saving: 40–60% on the initial creation phase.

GitHub Copilot: Intelligent autocomplete in the IDE to accelerate test writing
Cursor: AI IDE that understands your framework context and generates consistent code
Playwright Codegen + AI: Record interactions, AI cleans and structures them

4. Intelligent Test Prioritisation

ML can analyse execution history and identify which tests are most likely to detect a regression for a given code change. Result: run 20% of tests and detect 80% of bugs. Tools like Launchable and Saucelabs offer this feature.

5. Anomaly Detection in Data

For data-intensive applications, anomaly detection algorithms (Isolation Forest, LSTM) allow automatically detecting outliers in test results, even without pre-defined thresholds.

What AI Does Not Replace (Yet)

Understanding business context and priorities
Defining the test strategy (what to test, with what priority)
Root cause analysis during complex failures
Communication with product and development teams
Validating acceptance specifications with stakeholders

Master AI Applied to QA

Our Generative AI & QA Automation course teaches you to use LLMs, auto-healing and AI tools to multiply your productivity.

View the Generative AI course