Exploratory Testing with AI Agents

Traditional testing checks what you already know. Exploratory testing finds what you don't. Certyn's exploratory agent navigates your app like a curious user — clicking, typing, and investigating — to discover behaviors, edge cases, and bugs that scripted tests miss.

How It Works

You give the agent a goal, and it figures out the rest:

"Explore the settings page and check if all configuration options work correctly."

The agent:

1. Navigates to the settings area
2. Systematically interacts with every control
3. Records observations about anything unusual
4. Flags suspicious behavior for bug reproduction
5. Reports a session summary

No scripts. No selectors. Just a goal and a browser.

When to Use Exploratory Testing

New features

After shipping a feature, run an exploratory session to find edge cases the developers missed:

"Explore the new dashboard filtering feature. Test different filter combinations and verify results are consistent."

Unfamiliar areas

When you don't know an area well enough to write specific tests:

"Explore the admin panel and document what each section does. Note any inconsistencies or errors."

Pre-release sweeps

Before a major release, let the explorer scan broadly:

"Explore the entire application. Focus on navigation, form submissions, and error states."

After refactoring

When code changes don't have obvious test cases:

"Explore the checkout flow after the payment provider migration. Verify all payment methods still work."

Quick Tests vs. Exploratory Sessions

Certyn offers two modes for ad-hoc testing:

Quick test

A one-liner for fast, focused checks:

"Check if the forgot password flow sends a reset email."

Best for: verifying a single behavior, quick sanity checks.

Exploratory session

A goal with optional context and scenario:

"Explore the user profile section. Focus on: editing profile fields, changing avatar, updating notification preferences."

Best for: systematic discovery, broader coverage, unknown areas.

Both use the same exploratory agent. Quick tests are just shorter explorations.

What the Agent Records

Observations

As the explorer navigates, it records observations — findings that aren't necessarily bugs but are worth noting:

• Inconsistent UI patterns
• Slow-loading pages
• Missing validation
• Confusing UX flows
• Edge cases that behave unexpectedly

Observations are stored in your project and can be promoted to test cases or bug tickets.

Bug reproduction requests

When the explorer finds something that looks like a real bug, it doesn't just note it — it automatically spawns a bug reproducer agent to verify the issue is reproducible. More on this in the bug reproduction article.

Session summary

Every exploratory session ends with a summary: what was explored, what was found, and what needs attention.

Scoping Exploration

Broad exploration

"Explore the entire application."

Good for initial discovery. The agent will try to cover as much ground as possible within its time limit (15 minutes).

Area-focused

"Explore the billing and payments section."

The agent focuses on a specific area and goes deeper.

Scenario-driven

"Explore what happens when a user has no payment method on file and tries to upgrade their plan."

The agent follows a specific scenario path while still exploring around it.

Tips

• Start with broad exploration on new projects, then narrow down to specific areas
• Review observations regularly — they're your backlog of potential test cases and bugs
• Use exploratory testing alongside scripted tests — they complement each other
• The explorer's 15-minute time limit encourages focused goals
• Combine with device emulation and network simulation for realistic conditions
• Run exploratory sessions after every significant deploy to catch unexpected regressions