Prompt engineering

Use prompt engineering when the main problem is evidence rules, reusable prompt workflows, or verification before delivery.

Use this page when the main problem is how to ask the AI: which sources it may use, how to structure the prompt, and how to check the answer before you send it.

Prerequisites

Before you start

• Know what a good answer looks like for the task.
• Know how you will check whether the answer is correct.
• Have a draft prompt or a clear prompting problem to fix.

Scope

When to use prompt engineering

• Use this page when the main problem is the prompt itself, not product behavior, system orchestration, or API wiring.
• Use it when you need to decide where the AI may get facts.
• Use it when you need a repeatable prompt workflow for regular work.
• Use it when you need to check the answer before you send it.

Use another topic when the main problem is elsewhere:

• AI assistants — The main problem is how a chat product behaves.
• Agents and workflows — The main problem is how a multi-step system is routed, coordinated, or given memory.
• API integrations and tools — The main problem is tool calling, SDK usage, web access, or connecting the model to another system.

Continue with the right resource

Choose the resource that matches the next action, not just the content type.

Guides

Use these when you already have a draft answer and need a stricter review before you send it.

• Check claims with Chain-of-Verification Use this when you want to verify each important claim in a draft answer before sending it.
• Review wording with the Semantic Accuracy Gate Use this when the answer sounds right but may still be vague, overstated, or technically sloppy.

Policies

Use these when you want the AI to follow fixed rules about sources and final checks.

• Allow facts only from uploaded files Use this when the answer must stay inside the files you provided.
• Require cited public sources for facts Use this when factual claims must come from cited public sources.
• Require claim checks before the final answer Use this when the AI must verify important claims before it returns the final answer.
• Require a wording check before the final answer Use this when the AI must check wording, precision, and technical accuracy before delivery.
• Require web checks for current or niche claims Use this when the answer depends on recent or specialized public information.

Prompt templates

Use these reusable prompt templates when you want the same source rules and checking steps inside the prompt itself.

• System prompt: facts only from uploaded files Use this when the AI must not bring in facts from anywhere else.
• System prompt: cited public sources required Use this when factual claims must come from cited public sources.
• System prompt: claim checks before the final answer Use this when you want the AI to run a claim-checking pass before it answers.
• System prompt: wording review before the final answer Use this when you want the AI to check wording, scope, and precision before delivery.
• System prompt: web verification and citations for current topics Use this when the answer depends on current or specialized public information.
• Prompt template: web verification Use this when you want a step-by-step web verification workflow.
• Prompt template: citation format Use this when the final answer must present citations in a fixed structure.
• Prompt template: semantic accuracy review Use this when you want a reusable workflow for wording and technical accuracy review.

Reference

Use these when you want the background behind the checking methods used on this page.

• Verification techniques Overview of the checking methods used across this topic.
• Chain-of-Verification Background on how Chain-of-Verification works.

Articles

Use these when you want longer explanations behind prompt workflows, evidence boundaries, and the right ChatGPT operating mode for the task.

• Prompt Engineering Guide for Daily Work Long-form explanation of the prompt workflow used across this topic.
• Using ChatGPT Effectively at Work Practical guide to choosing the right ChatGPT mode, tool, memory setting, and project structure for each type of task.