AI-Generated Report Fact-Check Workflow
Systematically verify AI-generated reports before delivering to clients or publishing.
The 6-stage verification process
Each stage has defined inputs, AI instructions, expected outputs, and clear criteria for moving to the next stage.
Prepare
Gather content, set context, choose verification depth, select your AI model.
Extract
Extract every factual claim, classify by type, flag suspicious patterns using 6 fabrication categories.
Verify
Check flagged claims against evidence. Assess source quality using a 5-tier hierarchy.
Assess
Calculate weighted confidence score across 6 dimensions. Determine publishability.
Challenge
Adversarial review, citation trap scanner, and publication blocker check.
Report
Synthesize findings into a professional verification report with clear recommendation.
What’s in the workflow
Process architecture, not prompt wording
6-stage methodology with defined inputs, outputs, and decision logic at each step. Valuable even if every prompt changes.
3 human decision checkpoints
AI assists with analysis — you make the decisions. Clear guidance on what to review and when to override.
6 AI fabrication categories
Statistical invention, source fabrication, quote fabrication, temporal displacement, causal inflation, confabulated context.
Weighted confidence scoring
30% accuracy + 5x14% dimensions = objective publishability assessment with clear thresholds.
Built-in adversarial review
Red-team stage that stress-tests your content from a hostile reviewer perspective. Citation trap scanner included.
3 verification modes
Quick (30 min), Standard (1–2 hr), Deep (2–4 hr) — choose based on risk tolerance and time available.
Copy/paste execution
Every AI instruction is ready to paste. Fill in your content, run the prompt, review the output, move to next stage.
Complete package
START HERE guide, workflow map, 6 step cards, output templates, worked example, recovery guide, quick reference.
Who this is for
- Consultants delivering AI-assisted research to clients
- Content teams publishing AI-generated articles or reports
- Analysts using AI for internal or client-facing deliverables
- Researchers validating AI-produced literature reviews
- Anyone whose reputation depends on AI output accuracy
Works with your AI
ChatGPT (GPT-4o or later)
Claude (3.5 or later)
Gemini (Pro or later)
Any capable AI model
No API costs. No KaryoWorks account required. Use your own AI.
Pricing
Personal
$39
one-time
- Complete 6-stage workflow
- All AI instructions and templates
- Worked example
- Recovery guide
- Quick reference
- Prompt modification rights
Professional
$59
one-time
- Everything in Personal
- Use in client-facing work
- Email support
Team
$119
one-time
- Everything in Professional
- Up to 5 users
- Shared modifications
Frequently asked questions
- How is this different from the AI Output Verification Prompt Companion?
- The Prompt Companion ($29) gives you 15 individual prompts you can use standalone. This Workflow ($39) gives you the structured process — the sequence, decision logic, human checkpoints, branching, and output templates that tell you when and how to use those prompts. Think of it as the operating procedure vs. the individual tools.
- Do I need a paid AI subscription?
- No. This workflow works with free tiers of ChatGPT, Claude, and Gemini. A paid tier gives longer context windows (helpful for long documents) but is not required.
- What if the AI models change?
- The workflow's value is in the process — the decomposition, sequencing, verification methodology, and decision logic. Individual prompts may need adjustment as models evolve, but the process architecture remains sound. We provide updates for prompt adjustments.
- Can I modify the workflow for my domain?
- Yes. Your license includes modification rights for personal use. The workflow is designed to be adapted — the methodology is domain-agnostic, and prompts include clear variables you can customize.
- How is this different from the full AI Research Verification System ($99)?
- The System includes the complete methodology with scoring spreadsheets, evidence matrices, report templates, implementation checklists, and worked examples. This workflow is the executable AI process extracted from that system — structured for step-by-step AI execution.