20 practical guides
AI Literacy and Prompting Guides
Learn how generative AI produces output, how to write useful instructions, and how to build evidence, review, privacy, and cost controls around real tasks.
Built for people using generative AI in school, work, and personal projects who need concrete methods rather than prompt tricks.
How language models write
Understand tokens, next-token prediction, context, and sampling so you can interpret what an AI response can and cannot establish.
Four-part prompt structure
Write prompts with enough background, a precise task, useful boundaries, and an output shape that supports the next action.
Zero-shot and few-shot prompts
Choose between direct instructions and worked examples by testing ambiguity, edge cases, consistency, and the cost of a longer prompt.
Reusable prompt templates
Turn a successful one-off prompt into a documented template with named inputs, safeguards, tests, and clear ownership.
Reliable AI JSON
Define a narrow schema, request constrained output, validate every response, and recover safely when fields are missing or invalid.
Find AI hallucinations
Break AI output into checkable claims, compare each claim with authoritative evidence, and correct the record without preserving invented details.
Check AI citations
Verify that every AI-provided source is real, authoritative, current, and directly supports the claim attached to it.
Manage long AI context
Organize large inputs, retrieve relevant passages, control instruction placement, and test whether the model can find what matters.
Summarize long documents
Choose a summary purpose, preserve important qualifications, link statements to source locations, and audit omissions before sharing.
Extract document data
Define fields, retain evidence coordinates, validate values and tables, and send ambiguous documents to a human review queue.
Compare documents with AI
Align sections, classify meaningful changes, quote both versions, and verify high-impact differences before acting on them.
Analyze images with AI
Prepare visual inputs, ask bounded questions, preserve location evidence, and verify text, counts, and high-impact interpretations.
Better image prompts
Describe subject, composition, environment, visual treatment, and constraints, then revise one visible problem at a time.
Choose an AI model
Translate work requirements into capability, quality, latency, cost, privacy, and operational tests before selecting a model.
Control AI token cost
Measure input and output tokens, remove waste, cap responses, route tasks, and evaluate cost per accepted result instead of per request.
AI evaluation rubrics
Convert a task definition into observable criteria, severity levels, examples, reviewer guidance, and a repeatable test set.
Protect sensitive AI data
Classify data, minimize inputs, verify provider terms and settings, control access, and handle logs and outputs as sensitive records.
Brainstorm, then verify
Use AI freely for options, then switch modes, define claims, gather independent evidence, and approve facts through a separate review.
Workplace AI policy
Define approved uses, prohibited data and actions, review duties, incident reporting, ownership, and a practical update process.
Human review for AI
Place accountable reviewers at meaningful decision points with evidence, authority, time, criteria, and a clear escalation path.