Software Testing Research From TestMu AI

Original data from our calls and surveys with engineering and QA teams and from our nightly benchmark of AI testing agents, published with the method and the date behind each figure.

Read the 2026 Research

Two TestMu AI reports: AI Agent Testing Statistics from 2026, where 27% of 598 teams had no dependable way to test their AI agent, and the 2023 Future of Quality Assurance survey of 1,600+ QA respondents.

4 Research Reports on AI in Quality Engineering and QA

Each report shows its method and date. The AI agent testing research is refreshed each quarter, and the Verification Bench runs nightly.

See analyst reports on TestMu AI

AI IN QE 2026

How Teams Verify AI-Written Code and Test AI Agents

The State of AI in Quality Engineering 2026 survey of testers, developers and QA leaders, with every chart, its data and the full method.

Read the 2026 Survey Report

81% have merged AI code they didn't fully understand
37% of AI agent builders have no systematic evals
Charts with data tables, CSV copy and a citation

Chart from the State of AI in Quality Engineering 2026 survey: 81% of teams have merged AI-written code or tests they did not fully understand, 19% regularly, 33% occasionally and 28% rarely.

AI AGENT TESTING 2026

AI Agent Testing Problems Raised on 834 Recorded Calls

Buyer research from calls with engineering and QA teams, 31 July to 31 October 2026, with every testing problem ranked and the full method.

Read the 2026 Research

27% of 598 teams cannot dependably test their AI agent
19 recurring testing problems ranked by share of teams
Citation, embed code and CC BY 4.0 reuse terms

TestMu AI research chart: 27% of 598 engineering and QA teams have no dependable way to test their AI agent; of calls in which a team raised a problem (273, 280 and 229), 21%, 26% and 25% raised AI agent testing in August, September and October 2026.

VERIFICATION BENCH

Verification Bench Results for Kane CLI and Coding Agents

TestMu AI's nightly benchmark of Kane CLI against Claude Code and Codex driving Playwright MCP, with Kane CLI's PRISM 2026-Q4 scores.

See the Verification Bench

Kane CLI, Claude Code and Codex on 25 live workflows
False passes: 0.8% Kane CLI, 16.8% Claude Code, 13% Codex
Kane CLI's PRISM 2026-Q4 scores and the stated limits

Verification Bench card, October 2026 nightly runs: false pass rate as a share of all runs, Kane CLI 0.8%, Claude Code with Playwright MCP 16.8%, Codex with Playwright MCP 13%.

FUTURE OF QA 2023

Future of QA Survey of 1,600+ QA Respondents

The 2023 survey report on the trends, challenges and innovations shaping quality assurance, from 1,600+ QA respondents.

Read the 2023 Survey Report

Card for the Future of Quality Assurance survey report, published in 2023 with 1,600+ QA respondents.

TestMu AI (Formerly LambdaTest)/Research

TestMu AI for Enterprise

Get access to solutions built on enterprise-grade
security, privacy, & compliance

Advanced access controls
Advanced data retention rules
Advanced Local Testing
Premium Support options
Early access to beta features
Private Slack Channel
Unlimited Manual Accessibility DevTools Tests