Anthropic — Structured prompts tuned for Claude's careful reasoning and long-context strengths.
Use Claude-optimized prompts for long-document analysis, nuanced writing, careful multi-step reasoning, coding with large context, and tasks where tone and constraints genuinely matter.
Strong at working across large documents or codebases without losing track of earlier details.
Nuanced writing with good instruction-following on tone, constraints, and audience.
Reasons well step by step when a prompt asks it to show its work.
Will ask clarifying questions when prompted to, rather than guessing silently.
Bad prompt:
“Summarize this document.”
Structured:
Role: You are a research analyst. Action: Read the attached report and produce a 5-bullet executive summary. Context: Audience is a non-technical VP. Prioritize business impact over methodology. Example: Flag any claims that lack supporting data, similar to how an analyst would annotate a memo before forwarding it.
A 5-bullet summary a VP could act on without a follow-up question — business impact first, methodology skipped unless it changes the conclusion, and any unsupported claims clearly flagged.
Structured prompts tuned for GPT's conversational, general-purpose strengths.
Structured prompts tuned for Gemini's multimodal and research strengths.