Real Information – Artificial Intelligence: Prompting AI, Part 2

By Arthur Weiss

Editor’s note: This is part of a series covering what’s new and, more importantly, how infopreneurs can maximize the benefits of AI tools. 

In the last post, I gave a guide to the basic principles of prompting – knowing your objective and recognizing how prompts can bias responses. This time, we explore best practices and a practical framework for effective prompting.   

There are several frameworks used to create effective prompts – for example:

• CoT (Chain of Thought): Instructs the AI tool to reason step-by-step before answering; it is great for problem solving with a series of prompts, but it doesn’t specify the tone or output format and much more.
• RACE (Role, Action, Context, Examples) asks the prompter to describe the Role taken, the desired Action, the Context, and Examples. It’s easy to use and, with its focus on role and context, it can be useful.
• CARE (Context, Action, Refine/Clarify, Examples): Give the context, describe the action wanted, clarify, and give examples. For instance:
  We’re organizing a symposium for AIIP. Can you assist us in creating a targeted advertising campaign that emphasizes how we support infopreneurs. Our desired outcome is to increase awareness of AIIP and get sign-ups to the symposium. An example of a similar initiative was … and so on.
• CREATE (Character, Request, Examples, Adjustments, Type, Extras). It includes all the above, and the reason I like it is that it forces you to consider exactly what you want and additional specifics including output formats not provided with the CARE and RACE approaches.

CREATE has been attributed to Dave Birss, AI consultant and author. Birss claims to have developed this approach shortly after ChatGPT was launched in November 2022. Birss describes it in The Prompt Guide, a transcript for a LinkedIn learning course.

So, what is the CREATE framework? It’s quite simple but allows considerable complexity and refinement. (Note: It’s a framework – sometimes you won’t use all the steps.)

C = Character. What’s the AI’s role? How do you want it to answer you? Birss writes that it could be an expert, a celebrity, an artist, or an idiot. Make it clear with an introductory statement like:  You are an 18-year-old student planning your first budget backpacking trip across Europe. You have never traveled abroad before and are nervous about costs and logistics.

The AI should give an answer that is different from the answer to: You are a luxury travel consultant with 20 years’ experience arranging high-end bespoke trips for wealthy clients.

R = Request. This is the task you need done, or the question you want answered. The clearer you are about your objective, the better the chance of getting something useful.

I want you to design a detailed 5-day travel itinerary for Italy. I’ll be starting in Rome and I’d like to include Florence and Venice. I have a budget of $1000, I need to travel by train, and I want a mix of sightseeing and authentic local food experiences. If I need extra funds, ask me before adding expensive activities.

E = Examples. You may want to give more specific ideas on what’s wanted – perhaps with a list of information.

As I want this trip to be memorable, please don’t give me a generic tourist package that feels like a rushed coach tour. I’d love an itinerary that feels more like a Lonely Planet insider’s guide, with hidden gems, food markets, and neighborhood walks.

A = Adjustments. This is where AI can really shine – you can refine the prompt and make it precise and part of a conversation.

I gave you three cities to include, but you don’t need to pack every single landmark into the schedule; just point out what would make sense for a balanced trip. If you think I’ve left out a must-see experience, ask me if I’d like to add it before finalizing the itinerary.

T = Type. What format do you want? A 5-page report, a Word document, a table, bullet points, or a day-by-day travel schedule with timings?

Provide the plan as a 5-day itinerary in a table format, with columns for morning, afternoon, and evening. Include travel times, estimated costs, and recommended food stops.

E = Extras or special instructions. This can be a request to explain the thinking, use academic sources only, or ask for information if not everything has been explained to the AI to enable dealing with the request, e.g., “Ask clarifying questions before answering.”

Recommend only restaurants and attractions with strong reviews from reliable travel sites (e.g., TripAdvisor, Lonely Planet, Michelin Guide). Ask clarifying questions if my travel preferences (such as budget flexibility or interest in museums) aren’t clear before you finalize the itinerary.

In the next post, we put this framework into action and explore how AI can support deeper research and more complex tasks. (Full disclosure: I asked ChatGPT to suggest scenarios, and it came up with the travel example, Italy, Lonely Planet, and similar references.)

Arthur Weiss has been an infopreneur for almost 30 years. He founded AWARE in 1995 after a career at the business information company Dun & Bradstreet. He specializes in competitive and marketing intelligence using open sources (OSINT). Recently, he has pivoted to new areas, including exploring how AI tools can support infopreneurs. His latest insights can be read in International Marketing & Competitive Intelligence and Computers in Libraries magazines. He may be contacted at a.weiss@aware.co.uk.

Arthur will be hosting his next “AI Exchange” virtual event for AIIP on Wed, Sep 11, 2025 11:00 AM CDT. Register here!