This item was not updated in last three versions of the Radar. Should it have appeared in one of the more recent editions, there is a good chance it remains pertinent. However, if the item dates back further, its relevance may have diminished and our current evaluation could vary. Regrettably, our capacity to consistently revisit items from past Radar editions is limited.
Trial
AI-driven test generation has moved from experimental to practical. 81% of development teams now report using AI in testing workflows, and dedicated tools like Qodo and Mabl have matured enough to justify Trial.
Why It Moved from Assess to Trial
When this entry was first written in January 2025, AI test generation was promising but unreliable. Since then:
- Mainstream adoption: The AI testing market reached $687M in 2025, projected to hit $3.8B by 2035 (Grand View Research, AI Testing Market Report). Teams are getting measurable ROI, though real savings typically emerge after 12–18 months of adoption and process integration.
- Dedicated tools matured: Qodo (formerly CodiumAI) now offers IDE-integrated test generation with context-aware coverage analysis. Mabl's "Test Creation Agents" build entire test suites from natural language descriptions. Self-healing test frameworks (Virtuoso, Perfecto) handle UI test maintenance. Diffblue Testing Agent (GA March 2026) demonstrated 81% average line coverage on eight real Java projects vs. 32% achieved by a developer iterating with an AI coding agent (benchmark data).
- Built into everyday tools: GitHub Copilot, Cursor, and Claude Code all generate tests as a standard workflow — "write tests for this function" is now a routine prompt.
What Works Well
- Boilerplate test generation: AI excels at generating test structure, setup/teardown, and repetitive assertion patterns
- Edge case brainstorming: "What edge cases should I test for this URL parser?" — AI suggests scenarios humans miss
- Test data generation: Realistic fixture data for complex domain objects
- Coverage gap identification: Tools like Qodo analyze existing tests and suggest what's missing
What to Watch Out For
- Tautological tests: AI sometimes generates tests that verify what the code does rather than what it should do
- Missing domain knowledge: Tests that don't understand business rules give false confidence
- The paradox: The most upvoted question at a 2026 testing conference was: "Does AI-generated code reduce the need for testing, or demand MORE?"
Key Characteristics
| Property | Value |
|---|---|
| Type | Practice / workflow |
| Adoption | 81% of dev teams using AI in testing (2026) |
| Standout tools | Qodo, Mabl, Applitools, built-in IDE features |
| Best for | Boilerplate, edge cases, coverage gaps |
| Risk area | Tautological tests, domain-blind assertions |