CCAO-F facts to memorise: 30 recall cards
Thirty facts that are pure recall — the ones you either know cold or lose a mark on. Cover the right-hand column and work down the list.
| Prompt | What you need to know |
|---|---|
| Exam format | 60 questions, 120 minutes, 720/1000 scaled to pass, credential valid 12 months. |
| Heaviest domain | Output Evaluation and Validation, 21%. Governance (15%) outweighs Prompting (14%). |
| The core principle | Claude drafts; you remain accountable for anything you send, file or decide. |
| Three evaluation questions | Is it accurate? Is it complete? Is it appropriate? |
| Hallucination signature | Specific figures, citations or quotations that were not in material you supplied. |
| Lowest-risk output | Transforming text you supplied — rewriting, reformatting, translating, summarising an attachment. |
| Permission to say 'I don't know' | One of the highest-value sentences you can add to a prompt. |
| Grounding instruction | "Use only the attached documents; do not use general knowledge." |
| Making output auditable | Require a citation or supporting quote for each claim. |
| Best-of-N check | Ask twice; where the answers diverge, that is what needs verifying. |
| Do techniques remove the need for review? | No. They lower the error rate. High stakes still need a human. |
| Verification, low stakes | Internal brainstorm or draft: read it and move on. Over-checking is also a wrong answer. |
| Verification, high stakes | Primary sources plus a second reviewer. |
| The eight high-risk domains | Legal, healthcare, insurance, financial advice, employment, housing and lending, academic assessment, journalism. |
| High-risk rule | Qualified human reviews and owns the output, and AI involvement is disclosed. Both, not either. |
| Projects | Instructions + uploaded knowledge + own chat history. For recurring work over stable material. |
| Research | Many searches across external sources, returned with citations. For current information. |
| Artifacts | Substantial output in its own panel — for things you keep, revise and share. |
| Connectors vs uploads | Connector when the material lives in a system and changes; upload when it is a fixed document. |
| Incognito chat | Not saved to history, never used for model improvement. |
| Project sharing | View or edit permissions — a Team and Enterprise capability. |
| Model tiers | Haiku fast and cheap · Sonnet the everyday default · Opus complex work · Fable hardest reasoning. |
| Consumer training default | Opt-in, off by default. If enabled, de-identified retention up to five years. |
| Commercial training default | Team, Enterprise and API content is not used to train models by default. |
| What plans do not change | GDPR, HIPAA and contractual confidentiality still apply to the data itself. |
| Generic output | Missing audience, purpose and specifics — not missing model capability. |
| Forgetting earlier instructions | Conversation too long. Fresh chat with a summary, or move the context into a Project. |
| Outdated claims from a Project | Remove the superseded document. Do not instruct Claude to ignore it. |
| Big multi-part request comes back thin | Decompose into ordered steps with a check between them. |
| Communicating value | Concrete about the benefit, explicit about what needs checking, honest about what it cannot reach. |
← Previous: Questions & answers — Domain 7 | Next: Four-week study plan →
Explore Other Certification Study Guides
Switch to another exam guide to prepare for your next Claude certification:
Claude Certified Architect — Foundations
Core architectural principles: agentic loops & orchestration, tool design & MCP, Claude Code workflows, prompt engineering, and context reliability.
Claude Certified Architect — Professional
Advanced enterprise architecture: solution design, system integration, evaluation frameworks, governance, safety & developer enablement.
Claude Certified Developer — Foundations
Build software on Claude: API integration, model selection, tools & MCP servers, agents vs workflows, security, safety & debugging.