6 Prompt Engineering Strategies To Get Better ChatGPT Outputs

Have you ever found yourself frustrated with ChatGPT’s responses? Or maybe you’ve wondered if AI is just another overhyped tech trend that falls short of its promises. You’re not alone.

In fact, despite ChatGPT’s 180 million users, the majority aren’t tapping into its full potential. The flood of generic, AI-generated content plastered across social media, websites, and marketing content is proof that most users are barely scratching the surface of what AI l can really do.

Unfortunately, most users think it’s a problem with AI when actually, it’s how users are interacting (prompts) that’s the problem. I have narrowed this down to 2 reasons:

1: Laziness: Remember the days when blog posts took several hours to write? Creating unique ad copy would take hours of carefully selecting powerful words to have the most effect. I do. Now it takes just 30 seconds.

People are so blindsided by the efficiency and speed of AI that quality control and emotion have been lost from the final product (content). I recommend you check out the video below on the paradox of efficiency ‘where something becomes so efficient it results in decreased effectiveness‘ as it applies to this point.

2: Lack of training: Have you taken any form of training for using AI models or prompt engineering? Not many have, that’s why people use ChatGPT once and give up – or settle for the mediocre garbage it produces. Learning how to use AI models effectively will improve not only your experience but the quality of outputs you receive.

I have been a prolific user of AI – even before the public craze of ChatGPT and have researched and tested 6 prompt engineering strategies to help you use ChatGPT to its full potential. Remember that final outputs should still be edited and improved – don’t be lazy. It’s also worth noting, that most of these strategies are designed for backend setups of GPTs, they can still be effective from a front-end users POV!

Strategy 1: Write Clear Instructions

Let’s start with the most obvious – but the most underutilized. Writing clear instructions. When using AI models like ChatGPT for marketing, the details of your instructions directly influence the quality and relevance of the output you receive. Clear prompts leave little room for interpretation, ensuring the output aligns perfectly with your expectations. Sometimes it might take a few minutes to write the instructions, but it’s better than multiple amendments to the responses you get.

Below are some different tactics you can use with examples:

Tactic 1: Include Details in Your Query

One of the most effective ways to obtain high-quality content from ChatGPT is by embedding all necessary details within your prompts. This means moving beyond vague requests and providing specific information that directs the AI towards producing tailored and relevant content. Detailed prompts help ChatGPT understand the context, target audience, desired tone, and specific elements you want to highlight, resulting in outputs that are both precise and actionable.

Example

• Less Effective Prompt: Write a blog post about email marketing.

• Improved Prompt: Draft a 700-word blog post on advanced email marketing techniques for e-commerce businesses. Focus on personalization strategies, automation tools, and A/B testing to optimize campaign performance. Include actionable tips and recent industry statistics to support each technique.

Why It Works

While the less effective prompt can be a starting point for a generic blog post, it lacks detailed information and statistics. The improved prompt provides specific details such as the target audience (e-commerce businesses), key focus areas (personalization, automation, A/B testing), and additional elements (actionable tips, industry statistics).

The detailed prompt provides a good outline that can be edited into a richer and more detailed blog post, targeted to a specific audience. Take a look at the amount of detail provided in the prompt and try it for yourself in ChatGPT:

“Draft a 700-word blog post on advanced email marketing techniques for e-commerce businesses. Focus on personalization strategies, automation tools, and A/B testing to optimize campaign performance. Include actionable tips and recent industry statistics to support each technique.”

Tactic 2: Ask the Model to Adopt a Persona

Every brand has a unique voice and its content should reflect that. Whether your brand is playful and casual or professional and authoritative, your content must reflect this persona consistently. Instructing ChatGPT to adopt a specific persona encourages the responses to match your brand’s tone and style.

Example

• Less Effective Prompt: Write a LinkedIn post introducing a new content management tool.

• Improved Prompt: As a knowledgeable and approachable content strategist, write a LinkedIn post introducing our new content management tool. Highlight its user-friendly interface, seamless integration with existing platforms, and robust analytics features. Use a professional yet friendly tone to engage marketing professionals looking to streamline their workflows.

Why It Works

The less detailed prompt lacks brand personality and details which would require heavy editing and adjustments to publish live. The detailed prompt specifies the persona (knowledgeable and approachable content strategist) and outlines key features to highlight (user-friendly interface, seamless integration, robust analytics). Additionally, it directs the tone to be professional yet friendly, which aligns with the brand voice – although some small edits would still be required.

Tactic 3: Use Delimiters to Clearly Indicate Distinct Parts of the Input

Using clear separators like quotation marks, headers, or specific labels helps ChatGPT recognize different sections of your request. This makes it easier for the AI to process each part accurately, ensuring that your instructions are followed precisely without any confusion.

Example

Prompt: Please summarize the following sections into key improvement areas:

Customer Feedback:

1. The website is slow to load, especially on mobile devices.
2. The checkout process is too complicated with too many steps.
3. Lack of detailed product descriptions makes it hard to make informed decisions.
4. Customer support response times need to be faster.

Prompt: For the following text entries, I would like you to replace anything between the delimiters ‘<’ and ‘>’ with an emoji:

Yesterday, I visited the <museum> with my <friends>. We enjoyed the <exhibits> and had lunch at the <cafe> inside the museum. The <staff> were very helpful and made our day memorable.

Prompt: I would like you to write a short essay about the book The Hitchhiker’s Guide to the Galaxy with the following sections:
[Summary]: Provide a concise summary of the book
[Character Analysis]: Analyze the character of Arthur Dent in the book
[Thematic Discussion]: Discuss the theme of dispossession in the book

Why It Works

Clearly structuring your prompts reduces the possibility of ChatGPT missing key information and allows for easier refinements for certain sections. It also allows the AI to address each section of your request appropriately, leading to a more effective outcome.

Prompt source.

Tactic 4: Specify the Steps Required to Complete a Task

When dealing with multiple steps in a request, outlining a sequence helps ChatGPT understand the order of operations and the specific actions required. It applies logic and encourages the AI to think through the different stages of the process, and once outlined, you can further expand on the individual steps a lot easier.

Example

• Prompt: Create a step-by-step guide for launching a new podcast series. Start with concept development, move on to equipment selection, then cover recording and editing, followed by publishing and promoting the episodes.

Why It Works

By specifying each step in the podcast launch process, the prompt guides ChatGPT to produce a detailed and organized guide. You will also notice that the format is easy to follow and read, exactly how a step-by-step guide should be.

Tactic 5: Provide Examples

Providing examples within your prompts can help ChatGPT understand the format, style, or type of content you expect. This technique, known as “few-shot” prompting, is especially useful when you want AI to mimic a particular style or follow to specific guidelines.

Example

• Prompt: Write a product description for our new wireless earbuds. Follow the style of the example below:
  
  Example: Experience unrivalled sound quality with our latest wireless earbuds. Designed for in-ear comfort and military-grade durability, they offer lossless connectivity and all-day battery life, making them perfect for on-the-go lifestyles.

You can also do this with links to landing pages or copying text from other websites, take a look at this prompt example:

• Prompt: Write a product description for our new over-ear headphones. Follow the same style as the example below but keep it short:
  
  Example: Here I just copied the product description from Apple’s AirPods Max on their website.

Why It Works

By providing an example product description, the prompt gives ChatGPT a clear template to follow which guides the AI in matching the the desired tone, structure, and level of detail, maintaining consistency across your product listings. Notice in the Apple example prompt, how the output uses similar wordings and descriptive, confident language.

Tactic 6: Specify the Desired Length of the Output

Defining the length of the desired output helps ChatGPT adjust responses to meet specific requirements, whether you need a summary or an in-depth essay. Clear length specifications ensure that the content is appropriately detailed and matches your required use case.

Example

• Prompt: Summarize our latest quarterly report in approximately 100 words, highlighting key achievements, revenue growth, and future projections.

Why It Works

By specifying the desired length (approximately 100 words) and the key points to cover (achievements, revenue growth, future projections), the prompt directs ChatGPT to produce a concise summary. Whereas if you asked for only a summary, with no desired length, you could get masses of information or not enough.

Putting These Tactics Into Action

Imagine you’re preparing a content calendar for the upcoming quarter. To create an effective content strategy with ChatGPT, you’d employ all six tactics under this strategy:

Prompt: Develop a detailed content calendar for our digital marketing agency for the next three months.

Details:

• Target Audience: Small to medium-sized businesses in the tech sector.
• Content Types: Blog posts, social media updates, email newsletters, and webinars.
• Key Themes: SEO strategies, social media marketing, email automation, and content creation tips.
• Goals: Increase website traffic by 20%, boost social media engagement by 30%, and grow the email subscriber list by 15%.

Tasks:

1. List four blog post ideas for each month, ensuring they align with the key themes.
2. Create a brief outline for each blog post.
3. Suggest two social media updates per week to promote each blog post.
4. Draft an email newsletter template for monthly distribution.
5. Plan one webinar topic per month, including key points to cover.

Length: Ensure the entire content calendar is presented in a clear, organized table format with concise descriptions for each entry.

Example: An example is attached, please use this as a reference.

Resulting Output:

You will notice that ChatGPT has followed every instruction perfectly (the email template was provided as text following the calendar). The image only shows the first 2 weeks as there is too much information to show in the images, but you can see how effective a refined and detailed prompt using all of the tactics above can be:

Strategy 2: Provide Reference Text

Context is a vital part of any form of content, and providing ChatGPT with relevant background information or reference materials ensures that the generated content is accurate and reduces the chances of errors. This is slightly different than providing example text to copy, as in this instance you are almost training the AI on materials you provide to improve every output of the conversation. It could be a product brochure, SOPs, or company reports, but supplying these materials helps the AI understand your brand better.

Here are some tactics to consider:

Tactic 1: Instruct the Model to Use a Reference Text

Directing ChatGPT to base its responses solely on provided reference material ensures that the content is accurate and contextually relevant. This tactic minimizes the chances of AI introducing unrelated or incorrect information, as it relies exclusively on the materials you supply.

Example

Prompt: Using the company’s mission statement below, outline three key benefits of our customer loyalty program.

Reference Text: Our mission is to deliver exceptional value through innovative solutions and unparalleled customer service.

Why It Works

By providing the mission statement, the prompt helps ChatGPT understand the core values and objectives of the company to generate benefits that are directly aligned with the company’s mission. Of course, the example result above is poor in my opinion, but this is just for explanation purposes, it can be refined and optimized by providing further instructions and more detailed notes.

Tactic 2: Instruct the Model to Include Citations from Reference Text

Asking ChatGPT to cite specific sections from the provided text adds a layer of credibility to the generated content. This tactic supports the assertions made in the response and makes it easier to verify the information.

Example

Prompt: Based on the provided product brochure, explain the advantages of our latest smartphone model and cite relevant sections.

Reference Text or File Upload: [Insert Product Brochure Text Here]

Why It Works

By requesting citations, the prompt tells ChatGPT to reference specific parts of the product brochure, which ensures that the advantages highlighted are accurate and directly supported by the provided material, improving the reliability and consistency of the content. You can see in the example response that ChatGPT cites the pages it took the information from too, which can help you fact-check if needed.

Putting These Tactics Into Action

Imagine you’re launching a new service aimed at improving customer engagement through personalized marketing strategies. To create service descriptions and marketing materials with ChatGPT, you’d employ both tactics under this strategy, in addition to some of the previous tactics I covered:

Prompt: Develop a detailed service description for our new Personalized Marketing Solutions.

Reference Text: At [Your Company], we specialize in creating customized marketing strategies that drive customer engagement and boost sales. Our services include data-driven insights, targeted advertising campaigns, and comprehensive analytics to ensure measurable results.

Tasks:

1. Service Overview: Describe the Personalized Marketing Solutions service, emphasizing its unique features.
2. Key Benefits: Outline three main benefits customers will gain from using this service.
3. Implementation Process: Explain the steps involved in implementing personalized marketing strategies for clients.

Length: Ensure the service description is approximately 500 words, structured with clear headings and bullet points where appropriate.”

Strategy 3: Split Complex Tasks into Simpler Subtasks

Attempting to manage large projects through ChatGPT can often lead to confusion, missed details, and errors. In addition to using the strategies above, breaking down large projects into smaller, manageable subtasks, can ensure better quality outputs. For example, developing an entire content strategy or a full marketing campaign requires many different stages, by simply prompting ChatGPT to do this as a whole, you will receive poor-quality results.

Tactic 1: Use Intent Classification to Identify the Most Relevant Instructions for a User Query

Before diving into the details, it’s important to identify the underlying intent of the user’s request. This helps to create instructions that are relevant and effective so that ChatGPT is more accurate.

Example

• Prompt: Classify the following query into its primary marketing task and secondary marketing tasks, then provide instructions for that category:
  
  I need to build an email marketing strategy for our new product launch.

Why It Works

By classifying the type of request, ChatGPT can understand the intent behind the request and suggest appropriate instructions. You and the AI model then have a clear view of the different parts of the project, allowing you to prompt each one and refine it as you go, ensuring better quality outputs with fewer errors.

Tactic 2: Summarize Dialogue that Requires Very Long Conversations

In long conversations, it’s easy for important details to get lost. By summarizing or filtering previous dialogue, you help ChatGPT maintain focus on the most relevant information, ensuring continuity and coherence in the responses. You can also instruct ChatGPT to retain those key points and disregard all other information from the conversation, which can refine outputs.

Example

• Prompt: Summarize the key points from the previous conversation to maintain context for the next question.

Why It Works

Summarizing previous dialogue helps ChatGPT retain important information without getting overwhelmed by the entire conversation history. This leads to more accurate and context-aware responses. It also helps if you come back to the conversation after some time, instead of searching for information, you can ask ChatGPT to recall the important points it noted.

Tactic 3: Summarize Long Documents and Construct a Full Summary

When dealing with lengthy documents, summarizing them all at once can be challenging as there are limitations to AI models. Breaking the document into smaller sections, summarizing each part, and then combining these summaries into another summary, creates an overview for further use, without missing key details.

Example

For this example, I used a PDF of the book Dracula just to show the process, however, the outputs would be slightly different when using marketing materials.

• Prompt: Summarize each chapter of the provided report individually, then compile these summaries into a complete overview.

Why It Works

This approach ensures that every part of a document is accurately summarized, which informs the final summary is error-free and accurate. It prevents important information from being missed if the entire document were to be summarized in a single step.

Putting These Tactics Into Action

Imagine you’re developing a content strategy for a new product launch. To ensure that every aspect is covered thoroughly and accurately, you’d employ all three tactics under this strategy, in addition to some of the previous ones I covered:

Detailed Prompt: Create a detailed content strategy for our upcoming smart home security system launch.

Tasks:

1. Intent Classification: Determine the primary goals of the content strategy.
2. Dialogue Management: Review and summarize the previous planning discussions to maintain context.
3. Document Summarization: Summarize the market research report, the product specifications, and the competitor analysis in separate sections, then compile these summaries into a cohesive strategy document.

Length: Ensure the final content strategy is comprehensive yet concise, approximately 2,000 words, with clear headings and actionable steps.

Resulting Output:

The output does not reflect accuracy here as I do not have existing conversations or documents to provide – this is just an example.

Strategy 4: Give Models Time to “Think”

In content marketing, the quality of your output can impact your audience’s engagement and trust. By giving ChatGPT time to “think” through a problem before providing an answer, you enable the AI to generate more accurate and thoughtful responses. This strategy involves encouraging the GPT model to reason through its answers so the content is accurate and meets your requirements.

Tactic 1: Encourage Reasoning Before Final Answer

Asking ChatGPT to outline its thought process before delivering the final answer ensures that the response is based on a thorough analysis of the relevant factors. This approach helps in producing better content, reducing the likelihood of unfactual and poor responses.

Example

Prompt: Before answering, list the key factors to consider when launching a new email marketing campaign for our eco-friendly products.

Why It Works

By asking ChatGPT to first identify the key factors, you ensure that the answer addresses all primary aspects of the campaign. When you know the AI can understand the important parts of materials, you gain confidence in making further requests.

Tactic 2: Use Inner Monologue to Structure Responses

Instructing ChatGPT to internally process information while only presenting the final summary or conclusion helps maintain clarity and focus in the response. This tactic is particularly useful when you want detailed reasoning to inform the answer without overwhelming the reader with the entire thought process.

Example

This example is just for demonstration purposes, no real campaign data has been provided for the model to analyze.

Prompt: Analyze the effectiveness of our last social media campaign internally and provide a summary of your findings without showing the detailed analysis.

Why It Works

This prompt directs ChatGPT to perform a thorough analysis internally but only share the key insights. It ensures that the summary is concise and relevant, making it easier for you to understand the outcomes without getting bogged down by the entire analytical process.

Tactic 3: Ask the Model if It Missed Anything on Previous Passes

After receiving an initial response, asking ChatGPT to review its previous answers for completeness can help identify and fill any gaps. This is good practice for outputs that require accuracy and factually correct data.

Example

This example does not use any data, it’s a generic output that ChatGPT made up, but shows the result of such a prompt.

Prompt: Review your previous summary of our marketing strategy and identify if there are any key areas you might have missed. Provide additions if necessary.

Why It Works

Requesting a review of the initial response prompts ChatGPT to reassess and improve the content output, making sure that key elements of the marketing strategy are covered. Be careful with this prompt as using it repetitively can cause the AI to backtrack and adjust original answers in an attempt to satisfy. I would recommend using this only once to quality-check an output.

Putting These Tactics Into Action

Imagine you’re evaluating the success of a recent email marketing campaign aimed at promoting your new line of eco-friendly products. To do a thorough analysis with ChatGPT, you’d use all three tactics under this strategy:

Detailed Prompt: Evaluate the success of our recent email marketing campaign for our eco-friendly product line.

Tasks:

1. Encourage Reasoning: Before providing your analysis, list the key factors that should be considered when evaluating an email marketing campaign’s effectiveness.
2. Use Inner Monologue: Analyze the campaign’s performance based on the factors listed and summarize your findings without detailing the entire analysis process.
3. Iterative Review: After your initial summary, review your analysis to check if any important aspects were overlooked and provide additional insights if necessary.

Resulting Output:

Strategy 5: Use Feedback to Refine Prompts

In prompt engineering, you can improve your results by analyzing ChatGPT’s own responses and using that information to refine your prompts. Assessing the quality of the AI’s output and revising the prompt based on that, you create a feedback loop within your interaction, continuously improving outputs. To clarify, this is about analyzing ChatGPT’s response and finding gaps or weaknesses. For instance, if the response doesn’t include enough detail, you provide a refined prompt that asks for more. You are effectively getting feedback from the initial response, learning from what’s missing, and improving your next prompt based on that.

Tactic 1: Review and Adjust Prompts Based on Output Quality

Evaluating the content that ChatGPT generates helps to identify areas where it may have misunderstood or failed to include important details. Adjusting the prompt to address those factors guides future outputs to be more complete and relevant. This process helps you refine your prompts going forward, leading to more accurate responses with less back and forth.

Example

• Initial Prompt: Write a summary of our latest marketing campaign.

• Refined Prompt: Provide a detailed summary of our latest marketing campaign, including the main strategies employed, key performance indicators (KPIs), and the outcomes achieved.

Why It Works

Specifying the need for detailed strategies, KPIs, and outcomes, guides ChatGPT to produce a more informative summary, improving upon the first response. This will encourage you (the user) to use feedback and refine prompts in further conversations/outputs.

Tactic 2: Ask ChatGPT to Evaluate Its Own Output

After generating a response, you can ask ChatGPT to review its own output and determine if anything is missing or incorrect. This internal feedback mechanism allows the model to identify gaps or errors and improve the quality of its next response. It may also pick up on incorrect/poor outputs without you having to highlight them.

Example

No data was provided for this example, therefore the responses will not be as effective.

Initial Prompt: Summarize our last product launch event.

Follow-Up Prompt: Did you miss anything in your summary of the product launch? If so, add more details about customer feedback and product performance.

Why It Works

By asking ChatGPT to review its own output, you prompt it to check for overlooked parts of the content. This refines outputs to include more detailed responses so that key metrics or information are included in the final version.

Tactic 3: Use A/B Testing for Prompt Variations

Creating multiple versions of a prompt and comparing the outputs to determine which version yields the best results helps identify the most effective phrasing, structure, and level of detail needed to get your desired response. Even by following most of these strategies, you still might not be satisfied with specific outputs – therefore personal testing and optimization are key.

Example

• Prompt A: Generate ideas for our spring marketing campaign.

• Prompt B: Generate ten creative and actionable ideas for our spring marketing campaign targeting millennials, focusing on social media engagement and influencer partnerships.

• Prompt C: Provide your top ideas for marketing ideas that can be used in Spring specifically.

• Prompt D: Ideate top marketing campaigns for our product range throughout springtime. The campaigns should focus on brand awareness and driving sales. Only provide a brief overview as we will need to drill down into the nuances of each one personally.

Why It Works

Testing different prompts against each other will not only guide ChatGPT to provide better outputs that match your requirements, but it will also improve your prompt writing for future conversations. For example, if you learn that certain words or phrases trigger better responses, you can use them more often.

Strategy 6: Control the Tone and Style of Responses

The tone and style of content content produced by ChatGPT, for me, is probably one of the most identifiable factors for AI content. That’s why it’s not only essential to edit the content but also to get ChatGPT to follow different tones and styles.

Tactic 1: Define the Desired Tone in the Prompt

Specify the tone you want ChatGPT to adopt in its responses, such as friendly, professional, casual, or authoritative.

Example

• Prompt: Write a friendly and engaging Facebook post announcing our new product launch, highlighting its unique features and inviting followers to join the live demo.

You can also take this a step further and apply a persona as I covered further up, or even add additional descriptions to the tone/style such as this one:

Prompt: Adopt a friendly tone as if you were talking to a dog and tell me what the weather is like.

Why It Works

By defining the tone you want, the output is going to align with your brand’s messaging style, more than a generic prompt would. Test this out and pay attention to the choice of words, emojis, and other factors that demonstrate the tone effectively.

Tactic 2: Provide Style Guidelines or Examples

Including style guidelines or examples within your prompts helps ChatGPT understand the specific formatting, language, and preferences you expect based on existing approved content. While this prompt tactic does improve outputs, it won’t be 100% perfect, so I still recommend checking and editing once you have finished generating the content.

Example

For this example, I used an existing blog post from AHrefs on ‘how to get search traffic without ranking‘. This example only has 300 words, but the greater the length – the more you will notice the similarities:

• Prompt: Create a 300-word blog post about the benefits of organic traffic. Follow the style of our previous post titled ‘How to Get Search Traffic Without Ranking for Anything,’ which is pasted below and uses a conversational tone, includes bullet points, and relevant statistics:

Why It Works

Referencing a previous post’s style provides a clear template for ChatGPT to copy, which helps to ensure that the new content reflects tone, structure, and formatting. You can try this with any materials – whether you want to paste the content directly or upload the file, the AI will analyze and attempt to replicate the style, tone, and formatting. Of course, if you combine ALL these strategies and tactics together, you should get some great results.

So there you have it, 6 prompting strategies that can be used to improve your ChatGPT conversations. I still remain adamant that ChatGPT should be there to assist, not to replace, and additionally, ALL content produced by AI should be thoroughly reviewed and edited. The editing is not a suggestion, but a must; no content should be copy/pasted directly into your marketing channels and published externally. Good luck!