Do you ever feel you’re not getting the most out of AI tools like ChatGPT? You type in your question or task, but the response just doesn’t quite hit the mark. Don’t worry, you’re not alone! The key to unlocking the real power of AI is writing effective prompts.
An AI prompt is simply the instruction or query you give to the AI model. With better prompting skills, you can get AI to write novels, code programs, analyze data, and much more! In this guide, I’ll share my top tips for crafting prompts that get you the best AI outputs.
What are AI Prompts and Why Do They Matter?
An AI prompt is the initial instruction, question, or task you give to an AI language model like ChatGPT. The prompt tells the AI what you need it to do. With a clear, well-crafted prompt, you can get the AI to generate amazing essays, stories, code, analysis and so much more.
But with a vague or poorly worded prompt, you’ll likely get an unhelpful or confused response from the AI. Think of it like a chef following a recipe – a precise recipe will yield great results, but vague instructions will confuse the chef. Prompting is the key skill you need to unleash the full capabilities of AI truly!
Be Clear About the Task
The first rule of great AI prompting is to be clear about what you want the AI to do. Do you need it to write a short poem on a specific theme? Explain a complex concept like quantum physics? Analyze a set of data and spot trends or insights? The clearer and more specific you can be about the task, the better the AI can meet your needs.
Being vague like “write about AI” will likely get you a generic, rambling response. However, prompting the AI with an exact goal like “write a 300-word blog post intro explaining what AI prompts are and why they’re important” gives the model clear direction on what to focus on. You wouldn’t tell a chef to “make some food” – you’d specify if you need an entree, appetizer, dessert, etc. Same with AI prompts!
Provide Helpful Context and Examples
AI models are incredibly knowledgeable, thanks to their broad training data. But they can’t read your mind for the specific context you’re working with. The more relevant context and examples you can provide upfront, the easier the AI can understand exactly what you want and tailor its output accordingly.
Let’s say you want the AI to compose a persuasive speech in favor of plastic bag bans to be delivered at your town council meeting. Giving context about the target audience (the town council), the key arguments (environmental impacts, etc.), and maybe even an example of another short persuasive speech can help steer the AI model in the right direction.
If you just said “Write a persuasive speech about banning plastic bags” with no additional context, the AI would have to make a lot of assumptions that may be incorrect.
Use Simple and Unambiguous Language
When communicating with an AI model, it’s best to use straightforward, uncomplicated language that cannot be interpreted in multiple ways. While humans can often discern the intended meaning from context and nuance, AI models are more literal.
Avoid unnecessarily complex words, vague terms like “stuff” or “thingy,” and idioms or common figures of speech the AI may not understand. Being as clear and direct as possible eliminates any room for misunderstanding or confusion.
For instance, rather than saying something like “whip me up a doozy of a poem about technology,” try “write a thought-provoking poem around 20 lines long that explores the impacts of modern technology.” The AI can then focus on delivering exactly what you asked without making problematic assumptions.
Break Down Complex Tasks Step-by-Step
For big, complicated tasks that involve multiple components, you may want to break them down into smaller, more manageable steps rather than giving one gigantic prompt all at once.
Let’s say your goal is for the AI to plan and write a full 30,000-word science fiction novel about humanity’s first contact with alien life. That’s a massive ask to fit into one prompt! A better approach could be:
- Have the AI create a basic plot outline and describe the main characters and their backgrounds.
- Once you’re happy with that output, you could instruct it to draft chapter 1 of the novel based on the outline and character info.
- After completing Chapter 1, ask it to write Chapter 2, and so on…
Taking it step-by-step makes the task less overwhelming for the AI. It also allows you to refine and adjust as you go along. If the AI misses the mark on an early step, you can easily course-correct on the next step before it compounds further.
Specify the Desired Output Format
AI language models like ChatGPT are versatile and can generate outputs beyond written text paragraphs. When giving your prompts, ensure you specify the exact format you’d like the AI’s output to be in.
Do you need code written in a programming language like Python or JavaScript? An essay or report in a particular style like MLA or APA format? Or perhaps you need the AI to create an image, diagram, or other visual output?
Let the AI know upfront what the desired output format is. If the output format isn’t explicitly stated, the model will likely default to a standard paragraph text in plain English. So if you need the AI to generate code, for example, you’d want to say something like “Write a Python function that does X…” or you may get a text description rather than actual code.
Be Willing to Iterate
Don’t get discouraged if you don’t achieve your desired result on your first attempt at prompting the AI. Crafting an effective, precise prompt that hits the mark is often an iterative process that requires back-and-forth refinement.
If the AI’s initial output in response to your prompt doesn’t meet your standards or expectations, that’s okay! The key is being willing to provide feedback to the AI by rephrasing your prompt, adding more details and examples, tweaking the instructions, or adjusting the tone.
You can even share the AI’s flawed output with it directly and say something like “Thanks for this first draft, but it doesn’t quite capture what I was looking for. Please try again, this time making sure to…” and list how the revised prompt should differ. The more you engage with the model, provide feedback, and iterate, the better the subsequent outputs will be. It’s a learning process, but one that leads to prompting mastery.
Set the Right Tone
The specific wording and phrasing you use in your prompt can influence the tone and style of the AI’s output. For example, using more conversational phrases like “tell me about…” or “explain to me…” in the prompt encourages the AI to respond in a similarly conversational, natural language tone.
On the other hand, using more formal prompts like “write a comprehensive report on…” or “provide an in-depth analysis of…” tells the AI to adopt a more scholarly, professional, and technical tone. You can even specify writing styles like “translate the following into a concise journalistic article” or “rephrase this as you would for an academic research paper.”
So think about the desired voice and tone for the AI’s output, and craft your prompt accordingly by matching its phrasing and vocabulary. This attention to tone from the start helps ensure the output needs little editing or revision later.
Check Your Assumptions
While AI models are powerful and trained on lots of data, it’s important to be mindful that they can sometimes make incorrect assumptions or have certain biases “baked in” based on the data they were trained on.
This can be particularly obvious when asking an AI model to provide opinions, analysis, or take stances on subjective or sensitive topics like politics, ethics, and social issues.
If you’re dealing with one of these types of nuanced, context-driven topics in your prompt, it’s wise to recognize the potential for bias or blind spots up front. You may even want to explicitly counteract that by adding instructions about aiming for neutrality, considering multiple perspectives, avoiding discrimination, etc.
Something like: “Write an objective overview of the debate around [controversial issue], being sure to represent all mainstream viewpoints fairly and without taking partisan stances.”
This extra framing can go a long way toward ensuring the AI’s output doesn’t amplify harmful assumptions or biases from its training data on sensitive subject matter. A little proactive prompting can yield a much more balanced, responsible result.
Mastering AI Prompts – Learn from Examples
One of the best ways to improve at crafting effective AI prompts is to study examples of prompts being iteratively refined and optimized. Let’s walk through a few instances to illustrate the tips from this post in action.
Example 1: Writing a Short Story
Initial Prompt:
“Write a short story about a magic ring.”
This open-ended prompt is too vague and lacks context, so the AI output would likely be unfocused. Let’s refine it:
Revised Prompt:
“Write a 2,000-word fantasy short story from the perspective of a powerful magic ring. Set it in a medieval kingdom, describing how the ring passes between owners and the chaos/intrigue it causes. Include imagery around its abilities and curses.”
By adding specifics around word count, genre, point-of-view, setting details, plot points, and stylistic direction, this focused prompt frames exactly what’s needed from the AI.
Example 2: Analyzing Data
Initial Prompt: “Can you look at this data and tell me what insights you find?”
Without seeing the actual data set or understanding the analysis goals, AI has zero context. Here’s an improved version:
Revised Prompt: “Analyze the attached sales data from Q1 2023 for our clothing e-commerce brand. Identify any meaningful trends in top-selling categories, pricing dynamics, regional performance, marketing campaign impacts, etc. Provide recommendations to improve sales based on your findings. The output should be a detailed 500-word analysis report.”
By including the specific data file(s), target timeframe, key areas to explore, desired output format, and length, this AI prompt is now clear.
Example 3: Writing Code
Initial Prompt:
“Can you build me an app?”
This extremely broad prompt lacks any context needed for coding. Here’s how to enhance it:
Revised Prompt: “Write a React Native app that connects to our company API (details attached) to pull the latest product inventory data. Display in a sortable table format, with fields for name, SKU, quantity, and price. Allow users to export data to CSV. Use React Navigation for routing. The code should have comments explaining the logic. Please send the full project repo when finished.”
With details around coding language, integrated data sources, UI requirements, specific features, deliverable format, and more, this revised prompt gives the AI a clear technical spec to follow.
By reviewing examples that showcase prompt iteration in action, you can better understand how to construct highly effective AI prompts for any use case. Start broad, then enhance step-by-step with key details until you’ve framed the perfect, contextual prompt to unlock AI’s full potential.
Conclusion
By following these tips on crafting clear, goal-oriented prompts with helpful context and examples, using direct language, breaking down complexity, and setting the right tone, you’ll master AI prompting in no time! Developing this skill allows you to unleash AI’s awesome potential across virtually any domain.
Don’t be discouraged if you don’t get perfect outputs immediately. Prompting AI effectively is a learning process that takes practice. But stick with it, apply the tips we covered, and you’ll be creating amazing AI-generated content, analysis, code, and more like a pro. The future of human-AI collaboration is hugely exciting. Prompting is your gateway to directing AI’s capabilities to their highest and best use. Have fun with it!
