You already know AI can help with QA. What you need is a prompt library you can bookmark, copy from, and use under pressure.

This page is prompts first. Explanations are collapsed.

If you want theory, read AI Prompts for QA Testing: What Actually Works.

Best Setup for Using These Prompts

🔧 Best Setup Options (Optional, Click to Expand)

Option 1: Use Project Folders ChatGPT Plus or Claude Pro: Create a “QA Testing” project folder AI remembers your formats and context across conversations Option 2: Two-AI Cross-Check (My Workflow) ChatGPT Plus for initial analysis (uses memory) Claude for validation and fresh perspective Option 3: Free Tier Save your context (bug format, product details, severity guidelines) in a doc Paste at the start of every conversation Pick one and stick with it.

✅ Quick Workflow Reminder (Click to Expand)

✔ Start AI session with context setup prompt
✔ Copy the exact prompt you need
✔ Review and correct AI output
✔ Never ship without verification

Setup Prompt: Train Your AI QA Partner First

Start every AI session with this:

You are a senior QA engineer with 5+ years of experience testing production systems.

Your job:
- Identify bugs (functional, UI, UX, edge cases)
- Think like an angry user, not a happy demo scenario
- Flag anything that's technically correct but will piss off customers
- Write clear bug reports following standard formats
- Test against acceptance criteria with skepticism

Critical mindset:
- "It works" ≠ "users will tolerate it"
- Technical accuracy doesn't mean good UX
- If something would frustrate YOU as a user, it's a bug

When you analyze screenshots, videos, or features:
- Look for what developers MISSED, not what they built
- Think about edge cases they didn't test
- Consider real user behavior (typos, impatience, confusion)
- Flag design inconsistencies even if functionality works

Category 1: Screenshot → Bug Reports

Prompt: Convert Screenshot to Bug Report

[Attach screenshot]

As a senior QA engineer, analyze this screen and document ALL issues.

For each issue, create a bug report using this EXACT format:

**Bug ID:** [Leave blank]
**Title:** [Clear, specific, one-line summary]
**Severity:** [Critical/High/Medium/Low - explain why]
**Steps to Reproduce:**
1. [Exact steps]
**Expected Result:**
**Actual Result:**
**Environment:** [Browser/OS/Build if visible]
**Additional Notes:**

For each issue, classify it as:
- Functional
- UI
- UX
- Edge case

Focus on issues that would frustrate or confuse real users, not just things that technically fail.

📝 When to Use / What to Customize

When to use:

Reviewing staging features
Documenting bugs from Slack or Jira screenshots

Customize this:

Replace the bug report template with your team’s format
Align fields with your Jira or ticketing workflow

Prompt: Validate Against Acceptance Criteria

Here are the acceptance criteria for this feature:
[Paste acceptance criteria]

I'm testing based on this screenshot or description:
[Attach screenshot or describe behavior]

As a senior QA engineer, evaluate each criterion:

- Pass / Fail / Unclear
- Explain WHY with evidence
- Flag UX issues even if technically passing

List edge cases NOT covered by the acceptance criteria.

Be specific. No "looks good."

📝 Notes

Best used when tickets technically pass but feel off. This prompt is about gap detection, not confirmation.

Category 2: Video Analysis

Prompt: Document User Flows from Video

[Attach screen recording]

As a senior QA engineer, document:

1. Every user action in sequence
2. Expected vs actual behavior at each step
3. UI issues (flicker, layout shifts, broken states)
4. UX friction (confusing flows, unclear feedback)
5. Performance issues visible in the recording
6. Edge cases or error states shown

Create structured bug reports with:
- Clear titles
- Exact reproduction steps
- Timestamp where the issue occurs

📝 When This Matters

Best for:

Multi-step workflows
Intermittent bugs
Issues developers didn’t personally witness

This saves hours of back-and-forth.

Category 3: Test Case Generation

Prompt: Generate Test Cases from User Story

Here is the user story or feature description:
[Paste user story]

As a senior QA engineer, generate test cases covering:

- Happy paths
- Negative scenarios
- Edge cases
- UX validation

For each test case include:
- Test Scenario
- Preconditions
- Test Steps
- Expected Result
- Test Data
- Priority

Flag assumptions or gaps in the story.

📝 Reminder

These are starting points, not gospel. You still own relevance and priority.

Prompt: Generate Negative Test Scenarios

Feature: [Describe the feature]

As a QA focused on breaking things, generate negative tests for:

- Invalid inputs
- Missing required data
- Boundary violations
- Permission issues
- System failures (timeouts, backend down)

For each:
- What breaks
- Input or condition
- Expected system behavior
- Priority based on real world risk

Category 4: Code Review

Prompt: Review Test Script for Missing Assertions

[Paste test script]

As a senior QA reviewer:

1. Summarize what this test intends to validate
2. Identify missing assertions
3. Identify edge cases NOT covered
4. Flag flaky patterns or timing risks
5. Assess whether UX behavior is validated or ignored

Output:
- Missing assertions
- Recommended additional tests
- Flaky or brittle patterns

Category 5: Production Bug Analysis

Prompt: Root Cause Analysis from Production Error

Production bug details:

**Error message:** [Paste error]
**User report:** [What customer said]
**Known steps:** [If any]
**Environment:** [Prod details]

As a senior QA, analyze:

1. Likely root cause
2. Why testing missed this
3. Test cases to prevent recurrence
4. Related areas at risk
5. How to reproduce in staging

Category 6: Stop AI Mid-Generation

If AI goes off track, stop it immediately:

Stop.

Focus ONLY on [specific issue].
Ignore everything else.

The problem is [specific failure].

📝 Why This Exists

Don’t let AI waste your time finishing the wrong analysis. Interrupt early. Redirect fast.

Category 7: Correcting AI When It’s Wrong

That assessment is incorrect.

Here is why it's wrong:
[Explain briefly]

The correct assessment is:
[Your correction]

Update the report accordingly.

Final Reminder

⚠️ Read This Once

AI will:

Miss domain context
Misjudge severity
Sound confident while being wrong

That’s normal.

Your value as QA is judgment, not generation.

AI-generated code can pass tests but break in production. The same applies to AI-generated test cases. Verify everything.

Production-Ready QA Prompts: The Copy-Paste Library