Prompts for AI - Definition and Examples for ChatGPT & Midjourney

AI models such as ChatGPT and Midjourney have significantly changed the generation of ideas and content. While AI models existed before, they could only write simple texts. The current models are able to generate complex and creative content that is often almost indistinguishable from human authors. For this to succeed, users have to use so-called prompts.

We explain what prompts are and how to write them so that AI models deliver impressive results. We also go into practice and show concrete examples of prompts that you can use for your business.

Prompts - Definition

In artificial intelligence (AI), the term "prompts" describes short instructions or text fragments given to an AI model to obtain a desired response or output. Prompts are like task statements that ask the model to generate or understand certain information. They serve as a starting point for generating ideas, texts, translations, and answers to questions, for example. 

AI models such as GPT-3 use prompts to produce contextual results by following input instructions. The quality and precision of AI output depend on the clarity and relevance of the prompts. In this regard, the choice of words and the structure of the prompts, among other things, influence how well the AI accomplishes the desired task. Prompts are thus the central element in the interaction with AI systems and significantly determine their performance and applicability.

Prompt Engineering - Meaning

"Prompt engineering" is the formulation of precise and effective instructions for AI models to achieve desired results. Here, the user designs the prompt so that the model understands the specific task and responds correctly. This requires, among other things, clear language, active phrases, short sentences, context consideration, and possibly iteration to achieve the desired behavior.

Through clever prompt design, users can therefore not only generate content, but also solve complex tasks without having to train the model beforehand.

Overall, Prompt Engineering enables precise control of AI systems to ensure useful and reliable output for a wide variety of applications.

prompts tips

Writing Prompts for ChatGPT - 8 Tips

ChatGPT is one of the most used AI models on the market. To help you create effective prompts with AI on the fly, follow these 8 tips:

  1. Formulate precisely

    Make sure your requests are clear and concise. Refrain from ambiguous or vague wording to avoid misunderstandings and get accurate answers.

    Good query: "Explain the key features of OpenAI's GPT-3 model in terms of its architecture and application areas." 

    Why it's good: The request is formulated clearly and precisely. It gives clear instructions and asks for a specific answer.

    Bad example: "Tell me all about GPT-3." 

    Why it's bad: The request is vague and imprecise. It lacks clear instructions on what exactly should be covered via GPT-3, which can lead to a general and possibly inadequate response.

  2. Provide context

    Give the prompts generator enough context to understand what the question is about. Describe the situation, topic, or background of the question to get informed responses.

    Good request: "Describe the impact of GPT-3 on machine translation and provide specific examples of successful applications in this area." 

    Why it's good: The request provides sufficient context by clarifying the topic (impact of GPT-3 on machine translation) and the type of answer desired (description and examples).

    Bad example: "Tell me about GPT-3." 

    Why it's bad: The request does not contain sufficient context. It is unclear exactly what information is desired about GPT-3, which may result in a superficial or inappropriate response.

  3. Ask specific questions

    Ask specific questions that require a detailed answer. Open-ended questions often lead to more comprehensive answers, but they may not be precise enough.

    Good question, "What are the advantages and disadvantages of using GPT-3 compared to previous AI models?" 

    Why it's good: The query asks a specific question that requires detailed analysis. It promotes a more comprehensive answer with clear comparisons.

    Bad example: bad inquiry, "What do you think about GPT-3?"
     
    Why it's bad: The question is too general and leaves room for arbitrary opinions, without clear direction. The answer could be superficial or subjective.

  4. Use exploratory questions

    Challenge the model to think creatively or consider alternative perspectives to get interesting and varied answers.

    Good inquiry: "Imagine GPT-3 being used in education. What might the interaction between students and the AI look like? Describe possible scenarios." 

    Why it's good: The query encourages the model to think creatively and consider alternative perspectives. This can lead to interesting and varied answers.

    Bad example: "Describe the functions of GPT-3." 
    Why it's bad: The request is too sober and does not encourage creative thinking. The answer might be superficial and boring.

  5. Use start of record

    Begin prompts with a statement to indicate the type of response you want. For example, "Please explain...", "Describe the effects of...", etc.

    Good request: "Please analyze the impact of GPT-3 on the content marketing space. Consider both positive and potentially negative impacts." 

    Why it's good: The request starts with a clear instruction of what is expected. This leads to a focused and structured response.

    Bad example: "GPT-3 and Content Marketing." 

    Why it's bad: The beginning of the sentence is too vague and does not give a clear direction. The answer might be unstructured or insufficient.

  6. Structure answers

    If you need step-by-step explanations or procedures, make that clear to the AI. Formulate specifications such as "Name the 5 phases of ...". 

    Good Inquiry: "List the main phases of machine learning and briefly explain how GPT-3 could be used in these phases." 

    Why it's good: The request gives clear guidelines on how the answer should be structured. This leads to an orderly and well-organized explanation.

    Bad example: "What can GPT-3 do in terms of machine learning?" 

    Why it's bad:
    The request is too unspecific and leaves open which aspects of machine learning should be addressed. The answer could be unstructured and superficial.

  7. Customize prompts

    If the first response from ChatGPT does not have the desired content, revise the prompt slightly or phrase it differently to get more precise results.

    Good request: "The previous answer was helpful, however, I need more information on the applicability of GPT-3 in medicine. Could you please elaborate on this in more detail?" 

    Why it's good: The request shows how to customize the original prompt to get more specific information.

    Bad example, "Why wasn't your last answer accurate?"
     
    Why it's bad: The request is too defensive and gives no clear indication of what should be changed in the prompt.

  8. Interact with answers

    Take advantage of the ability to interact with the AI's generated responses. Ask queries to clarify content and ask for details if the answer is unclear.

    Good query: "The answer to my previous question was a bit unclear. Could you please provide examples of the potential applications of GPT-3 in the field of environmental science?" 

    Why it's good: The query uses the ability to interact with AI generation to get additional information and clarify ambiguities.

    Bad example: "Do I understand this correctly?" 

    Why it's bad: The request is too vague and does not contain a clear direction for the interaction. It is unclear what exactly the questioner wants to clarify.

9 Prompts Examples for ChatGPT

To get a feel for how AI prompts for ChatGPT should be worded, we show concrete AI prompts examples:

3 ChatGPT AI Prompts Examples for Marketing

1. Develop three creative strategies for launching a new face lotion. The product has only natural ingredients. The target audience is men aged 30-50.

Design a comprehensive social media campaign for the launch of a new tequila. Include suggestions for ideal posting frequency and creative content ideas. The tequila comes from Mexico, but will be distributed in fancier bars in Germany at launch. 

chatgpt prompts marketing
chatgpt prompts marketing
chatgpt prompts marketing
chatgpt prompts marketing

3. develop 5 campaign slogans for an e-commerce store selling handmade rings that convey the company's focus on quality and sustainability.

3 ChatGPT AI Prompts Examples for Content Creation

1. submit 5 suggestions for unusual topics for Instagram stories that will excite followers of a vegan bakery.

chatgpt example content creation
chatgpt prompts content creation

2. Generate 5 blog post ideas that promote a travel agent's services and expertise.

3. develop 3 creative TikTok video scripts promoting a young startup's trekking products.

3 ChatGPT AI Prompts Examples for Web Development

1. create a navigation menu that is optimized for usability. The menu should be designed to take users to all pages of an employment lawyer's website in no more than three clicks. 

2. design 3 optimized contact forms to collect emails from B2C customers in the fitness industry.

3. create a draft landing page for a mobile gaming app that will appeal to younger users.

chatgpt example web development
chatgpt example web development

Writing Prompts for Midjourney - 4 Tips

To work with the AI Midjourney to create original images, you should pay attention to these 4 tips: 

1. Do not use too many requests

It is important that the prompt does not contain too many requests. A long list of requests can cause the bot to have difficulty fulfilling all the requests and produce an Image that is coherent. Instead, limit yourself to a few precise queries.

2. Choose words wisely

Choosing the right words and phrases is critical to writing a good prompt for AI images with Midjourney. Use precise terms that give the bot a clear idea of what is to be generated.

3. Experiment with formulations

There are no hard and fast rules for what a good prompt for Midjourney should look like, so you should experiment with different phrases and find out which ones work best that way. Try different word combinations and see what images the AI generates.

4. Leave room for creativity

Good prompts for AI images from Midjourney should not be too restrictive. Leave room for creativity and innovation by giving the bot enough freedom to create a unique and exciting image.

Tip: A good place to get inspiration for midjourney prompts are image descriptions on commercial image platforms. For example, the image below has the following description:

"Happy and smiling young successful business woman standing with colleague looking at smartphone in modern office and coworking space"

Businessman And Businesswoman Smiling Looking At The Phone stock photo and more pictures from Office - iStock (istockphoto.com)

5 Prompts for AI Images with Midjourney

1. A modern office building at night, lit from inside, with a clear sky and stars in the background.

2. A team of employees collaborating and working on a project, with laptops and papers on a large conference table.

3. A happy customer holding a product in his hands and smiling, with a bright and friendly background.

4. An innovative product presented on a clean and minimalist background, with clean lines and bright colors.

5. A group of people working together in a relaxed environment, with plants and natural light.

FAQ

What are prompts?

Prompts are clear instructions that help AI systems generate ideas or content. The inputs serve as a starting point for the artificial intelligence to solve targeted tasks and generate precise answers. The results feature a human-like writing style or realistic images, for example.

How do I write prompts?

Effective AI prompts require clear wording and an active speaking style. Users should avoid passive voice and nouns. Active verbs provide users with precise answers and short sentences increase comprehensibility.

What is Prompt Engineering?

Prompt engineering describes the precise design of prompts for AI models such as ChatGPT. Specific results can be achieved through targeted wording. This method makes it possible to obtain relevant answers to complex questions. In doing so, users should always adapt their requests to achieve optimal results.

Would you like to receive more detailed information on the use of AI? Our experts will be happy to advise you on the possible use of AI software in your company. Contact us at any time via the Contact form!

"
"
Jan Schäfer Avatar

Latest articles