This is some text inside of a div block.
Prompting modalities
Instructional prompts
Instructional prompts focus on completing literal tasks that don’t require much nuance or deep reasoning. These are straightforward asks where the model simply executes an action.
Examples:
- Categorizing a blog post into one of your predefined categories
- Extracting specific text from a document
- Returning live information about a topic
Analytical prompts
Analytical prompts guide the model to analyze, compare, or reason through larger data sets or complex content. They’re designed to ask the right questions so the model can draw insights or solve problems that involve multiple pieces of information.
Conversational prompts
Conversational prompts leverage the model’s interactive nature to produce more human-like, engaging content. By treating the exchange as a back-and-forth dialogue, you can refine tone, style, and clarity, resulting in copy that feels natural and polished.
Mapping modalities to specific tasks
- Live information
Uses an instructional prompt with models like Perplexity or Gemini 2.5 to fetch up-to-date data. - Data analysis
Applies an analytical prompt on a reasoning model to compare datasets or reason through complex content. - Data extraction
Employs an instructional prompt paired with a model that supports JSON output mode, pulling out structured data after analysis. - Content writing
Uses a conversational prompt with writing models like Claude Sonnet to generate richer, more human-sounding copy.
