Understanding AI System Prompts

A How-To Guide


When working with AI-powered tools like OnVerb, getting the best possible output often hinges on how effectively you communicate your needs to the AI. That’s where “AI System Prompts” come into play. This how-to guide will help you understand system prompts and how to utilise them to maximise the AI’s potential.

What Are AI System Prompts?

AI System prompts are detailed instructions or contexts you provide to the AI to shape its understanding and responses to your queries. Think of them as initial settings or a briefing that inform the AI about the specifics of the task you are engaging it with.

When you interact with different AI chatbots you generally don’t get to access the AI system prompt as this has a specific command specific to that chatbot, though some do allow a context message. This is where you would add your system prompt.

However, if you interact with a service like ChatGPT or Claude via an API (secure access via 2 systems talking to each other), you supply a AI system prompt as you initiate the first chat message. OnVerb offers simple management of system prompts, so you can have many different system prompts for your specific scenario. Maybe like a GPT, the extensions for ChatGPT, though simpler and easier to create many different system prompts, and switch between different AI models, such as Claude, Mistral Medium, Google PaLM2, OpenAI’s ChatGPT 3.5 Turbo, 4.5 Turbo, Dall-E 2 and Dall-E 3.

Why Are They Important?

  1. Precision: They help the AI understand the intricacies of your requirement.
  2. Relevance: They ensure that the output is tailored to your specific needs.
  3. Consistency: They provide a standard starting point for the AI, ensuring uniformity in the quality of responses.

How to Use AI System Prompts

1. Define Your Objective

Before you even begin to craft a prompt, be clear about what you want to achieve. Are you looking for creative content, technical advice, or problem-solving? Your objective will dictate how specific your AI system prompt should be.

2. Identify Key Details

What essential elements must the AI know to perform the task correctly? This could include:

  • Audience type (technical, general public, age group)
  • Tone of voice (formal, casual)
  • Purpose (inform, persuade, entertain)
  • Specific requirements (word count, format, key points)

3. Structure Your Prompt

A well-structured AI system prompt guides the AI through the context intricately. A basic structure may include:

  • Introduction: A brief of your ultimate goal.
  • Context: Additional information delivering deeper understanding.
  • Task: Clearly stated instructions.
  • Examples: Optionally include examples for the AI to mimic or draw inspiration from.

4. Test and Refine

Craft a prompt and use it to engage with the AI. Examine the output for relevance and quality. If it doesn’t meet your expectations, refine the prompt by tweaking its elements.

5. Save and Manage Prompts

Once you’re satisfied with a prompt, save it for future use. This saves you time and ensures consistency across multiple tasks.

Example:

Let’s say you’re creating a prompt for a recipe that considers dietary preferences.

  • Introduction: I’m looking for a vegetarian recipe.
  • Context: The recipe should avoid nuts and should be spicy.
  • Task: Please create a recipe for a spicy vegetarian lasagne.
  • Examples: Excluding ingredients: nuts. Including ingredients: chilli peppers, tomatoes, lasagne sheets, ricotta cheese.

Tips for Crafting Effective Prompts:

  • Be Specific: The more precise you are, the better the AI can tailor its responses.
  • Prioritise Clarity: Avoid ambiguous language.
  • Be Concise: Although detail is good, brevity helps the AI to stay focused on the most critical aspects of the task.
  • Use Natural Language: Write as if you’re instructing a human assistant.

By mastering AI system prompts, you unlock the full potential of your AI tool, ensuring that your outputs are not just high quality, but also highly personalised to your requirements.