The quality of an AI answer depends heavily on the quality of the prompt you write. This guide summarizes the most effective prompting techniques recommended by the official Anthropic and OpenAI documentation, with examples you can apply right away in ChatGPT or Claude.
1) Be clear and specific
State the task, audience, tone, length, and output format. Instead of “write about marketing,” say: “Write 5 short bullet points for a beginner about email marketing, in a practical tone.”
2) Give examples (few-shot)
Including one or two examples of the desired output raises accuracy a lot, because the model mimics the pattern and format you showed it.
3) Set a role and context
Assigning a role (“you are a precise copy editor”) and adding the needed context steers both style and content. In Claude use the System field; in ChatGPT use custom instructions.
4) Structure the prompt and ask for step-by-step reasoning
Separate parts with clear headings or tags, and ask the model to “think step by step” on complex problems. Anthropic also recommends structuring prompts with XML tags to separate instructions from data.
5) Iterate and refine
The first prompt is rarely the best. Review the output, adjust the instructions, and break long tasks into a sequence of prompts (prompt chaining).
Official sources
- Anthropic — Prompt engineering guide
- OpenAI — Prompt engineering guide
