NEW UPDATE

THE ART OF
PROMPT
ENGINEERING
ChatGPT
with

GPT-4
Update

Mid-May Update

NATHAN HUNTER
 New Book Launch

www.ChatGPTtrainings.com/MidJourney
11
BUILDING PERSONAS

Adopting a design thinking approach and focusing on users is essential in

today's fast-paced world. By putting users at the heart of the design
process, we can truly understand what they need and want. For instance,
when creating a mobile app, knowing the user's needs leads to a more
enjoyable and easy-to-use experience. In healthcare, focusing on users can
result in treatments that are easier to access and better suited for patients
with different needs. Sometimes, we can be so caught up in our own
perspectives that we might not see what others need.

Now, let's explore how we can use ChatGPT to tap into the power of this
user-focused model with personas.

In this chapter, we're going to look at a situation where we're writing a

company newsletter about adapting to AI tools.

To use a user-focused approach for this scenario, we should start by figuring

out who will be reading our content and create mini profiles for each type of
person. Doing this helps us understand their needs better and create
content that connects with everyone.

Here's an example of a prompt we could give ChatGPT to help us create

various personas for our company newsletter:

ChatGPT, can you help me come up with 3 different personas for our
internal company newsletter about getting used to AI tools? Our
company specializes in marketing and has a diverse workforce,
including creative designers, content writers, marketing strategists,
and data analysts. Please consider different roles within the
organization, their varying levels of technical expertise, and their
unique concerns and interests when it comes to using AI. Give me a
short description for each persona, including their job title,
background, and main interests or concerns about integrating AI into
their daily work.
ChatGPT has provided us with three well-defined personas that cater to our
diverse workforce: Sarah the Creative Designer, Alex the Content Writer, and
Tom the Data Analyst. Each persona has unique interests, concerns, and
backgrounds that are relevant to our user-centric approach. This allows us
to tailor our content to address their individual needs and expectations,
ultimately making our newsletter more engaging and relevant to our
readers.
Now that we have our personas, let's move on to using them effectively.

Using Personas for Ideation

Before we start writing our content, it's important to think about what each
persona would want to get out of the newsletter, the format it should be in,
and the language we should use. By doing this, we can ensure that our
content is tailored to our audience and resonates with each reader.

Let's explore how we can work with ChatGPT to consider the needs of each
persona when creating our newsletter.

First, we can use the personas to brainstorm what kind of information and
insights each reader might be looking for in the newsletter. For example,
Sarah might be interested in learning about new AI design tools, while Alex
might want to know how to create engaging content using AI, and Tom
could be curious about AI-powered data analysis techniques.
Here's an example of a prompt we could write for ChatGPT, to help us
brainstorm some ideas for our newsletter.

ChatGPT, please help me brainstorm ideas for our internal

company newsletter about adapting to AI tools, keeping in mind
the interests of our three personas:
(Personas)
We've received a range of ideas from ChatGPT for our internal company
newsletter, focusing on the unique interests and concerns of our three
personas: Sarah the Creative Designer, Alex the Content Writer, and Tom the
Data Analyst. By addressing each persona's specific needs, we can create
engaging and relevant content that truly connects with our audience.

In the output, ChatGPT also considers the best formats and language styles
for every persona, ensuring that we're speaking their language and
presenting information in the most accessible way. For example, Sarah
would appreciate visual examples and straightforward language, while Tom
would find more technical language with a focus on practical applications
appealing. By taking these preferences into account, we can craft a
newsletter that's not only informative but enjoyable for everyone.

Using personas to get feedback on our work.

Just like focus groups, which gather people together to discuss their
opinions and reactions to a product or idea, we can use our personas to
help ChatGPT give specific feedback on our content. Focus groups are great
for understanding how different people might perceive and interact with
something, allowing us to make improvements based on their thoughts and
feelings.
By giving ChatGPT our personas and the content we've made, we can get
feedback that takes into account the unique needs and preferences of each
persona. This helps us understand how different people might see our
content and make changes to improve it, similar to gathering insights from
a focus group.
There's a small challenge with ChatGPT – it has a character limit for each
interaction. To work around this issue, we can communicate with ChatGPT
in two parts. First, we'll give ChatGPT our newsletter draft using this
prompt:

Here's the current draft of our company's newsletter. We want to

use a user-centric approach when improving the newsletter.
I will start by giving you the newsletter and you will ask me for
the personas, so you can give user-specific feedback.
(Newsletter)
This forces ChatGPT to automatically move onto our personas. If we just
gave ChatGPT the newsletter, it might try to summarise it, feedback it, or go
rouge in another way.

After getting this response from ChatGPT, we can then give the personas for
user-specific feedback. For our use case, this is what ChatGPT replied:

By using this two-step approach, we can get feedback from ChatGPT that
considers the interests and concerns of our diverse audience, similar to
collecting insights from a focus group. This helps us improve our content
and make it more appealing and effective for everyone.
Exercise
Ready to put your newfound knowledge of personas to practical use? This
exercise will guide you through the process of creating personas with
ChatGPT, tailoring content to their unique needs and interests, and refining
your work based on feedback. This hands-on approach will help you truly
understand the power of personas in content creation. Let's dive in!

1) Crafting Personas
Imagine you are creating a blog post about the future of renewable energy.
Use ChatGPT to help you create at least three different personas that would
be interested in your blog post. Consider factors such as their jobs, their
interests, and why they would be reading your post.

2) Tailoring Content
For each persona, brainstorm what kind of information they would be
looking for in the blog post. What format should this information be in?
What language should you use? Write down your ideas for each persona.

3) Working with ChatGPT

Interact with ChatGPT to brainstorm ideas for your blog post tailored to
each persona. Remember to consider the unique interests and concerns of
each persona.

4) Review and Reflect

Compare the ideas you brainstormed with the suggestions provided by
ChatGPT. Do they align? Are there any new insights or ideas you gained from
ChatGPT's suggestions?

5) Creating Content and Getting Feedback

Draft a section of your blog post for one of your personas. Ask ChatGPT for
feedback, specifying the persona for which the content is intended. Reflect
on the feedback you received. How does it align with the needs and
preferences of your persona? What improvements can you make?

6) Iterate and Refine

Apply the feedback from ChatGPT to refine your draft. Repeat this process
for all your personas until you have a blog post that caters to each of your
reader's interests and concerns.
12
CHAIN PROMPTING

As we have seen throughout this book, ChatGPT is a fantastic AI language

model that helps generate human-like responses to prompts and questions.
But remember that it works by guessing the next best word based on the
context. While this is useful for many tasks, sometimes it can struggle with
more complex problems that need a deeper understanding or a more
organised approach.

Imagine solving a tough maths equation. Trying to jump straight to the

answer can be confusing and often leads to errors. Instead, we break the
problem into smaller, easier-to-handle steps and work our way through
each one to get the solution. This step-by-step approach can also be applied
to ChatGPT, making it more effective at tackling complex tasks and
generating better outputs.

This is where the idea of Chain Prompting comes in. Think of a chain made
up of many links, with each one connecting directly to the next. With
ChatGPT, we can create a specific workflow that includes a series of
prompts or questions. The AI then uses the output from each step as the
basis for the next one. This method lets ChatGPT build momentum and work
through larger or more complex workloads step by step.

Chain Prompting helps users guide ChatGPT through a process that slowly
refines the output. By breaking complex tasks into smaller steps, the AI can
generate more specific, customised, and overall better results. This
approach is really useful for creating content, like articles or blog posts,
where having a structured outline is key for making the content easy to read
and follow.
One practical use case for Chain Prompting is writing a blog post that
performs well in terms of SEO (Search Engine Optimisation). SEO is the
practice of optimising your website's content, so search engines like Google
are more likely to rank it higher in search results. For a blog post, good SEO
means increased visibility, more organic traffic, and ultimately, higher
chances of reaching your target audience.

In our case, the primary goal is to create a blog post that ranks well in
search results, rather than focusing solely on the content itself. This means
we need to consider the various elements that contribute to SEO
performance, like keywords, headlines, and structure, when crafting our
blog post.

To start, we need to think about the step-by-step process that we would

follow if we were to write a high-performing SEO blog post. The first few
steps would include identifying common questions related to the topic,
generating a list of popular keywords and phrases, choosing the most
relevant questions and keywords for the headline and subheadings, and
creating an outline for the post.

Once we have a clear idea of the steps involved, we can then turn each step
into a prompt for ChatGPT. This can be done manually or by asking ChatGPT
to build the prompts for us. In doing so, we create a chain prompt that
guides ChatGPT through the entire process of crafting a high-SEO-
performing blog post.
Here's an example of a 10-step chain prompt for writing a blog post with
great SEO performance:

1) ChatGPT, please provide a list of common questions related to [Topic].

2) Based on the provided list of questions, let's generate a list of popular
keywords and phrases associated with [Topic], focusing on high search
volume and low competition.
3) From the list of common questions and keywords, please choose the most
relevant ones and create a compelling and SEO-friendly headline and
subheadings for a blog post about [Topic], incorporating the main
keywords.
4) Now that we have the headline and subheadings, provide an outline for a
blog post about [Topic], with each section addressing one of the chosen
common questions.
5) With the outline in place, for each section of the [Topic] blog post, provide
supporting points or key messages that can be developed into informative
and engaging content, along with relevant statistics, facts, or examples.
6) Let's discuss the best practices in on-page SEO for a blog post about [Topic],
such as optimal keyword density, URL structure, meta descriptions, and
image alt text.
7) Now, generate the introduction for the blog post about [Topic],
incorporating the headline, main keywords, and a brief overview of the
content.
8) Thanks! Generate the content for the first section (Subheading 1) of the
[Topic] blog post, based on the supporting points, key messages, and SEO
best practices.

Repeat step 8 for each subheading or section, modifying the subheading

number accordingly.

9) Great work! Now, generate the conclusion for the blog post about [Topic],
summarising the main points and providing a call-to-action for the readers.

10) Finally, for each section of the [Topic] blog post, provide an image prompt
for a text-to-image AI tool in the following format: '[6-word description of
the image or visual], [4 adjectives further describing the visual separated by
commas], [the visual style (keep it consistent throughout)]'.
Here's an example of a 10-step chain prompt for writing a blog post with
great SEO performance:

Let’s walk through the first few steps of this 10 step chain prompt together,
so we get a better understanding of how this works. For my example my
topic will be dogs, so my first prompt will be:
“ChatGPT, please provide a list of common questions related to dogs.”

I then follow up directly with the next prompt in the current chain:
Based on the provided list of questions, let's generate a list of popular
keywords and phrases associated with dogs, focusing on high search volume
and low competition.”
“From the list of common questions and keywords, please choose the most
relevant ones and create a compelling and SEO-friendly headline and
subheadings for a blog post about dogs, incorporating the main keywords.”

Remember that ChatGPT has a context limit, which is the maximum amount
of text it can take into account when creating a reply. For GPT-3.5, the
context limit is around 1,400 words in English. If the chain prompt goes
beyond this limit, ChatGPT might not remember all of the previous steps,
which sometimes leads to ChatGPT filling in the blanks with what you could
have asked for instead of what you actually asked for. Going beyond 10
steps may be more difficult for ChatGPT to stay coherent.
Thankfully, GPT-4, the latest version of the AI model, has a significantly
larger context limit—4 times larger than GPT-3.5 for now, with another
model in beta testing that’s limit is 16 times larger than GPT-3.5’s. This
ability to handle extra context makes GPT-4 more suitable for Chain
Prompting, as it can handle more extensive and complex prompts without
losing track of the context. This allows you to create longer, more detailed
chain prompts, helping ChatGPT produce even more precise and well-
structured content, just like with our current example of writing SEO-
optimised blog posts.
Exercise

Ready to give Chain Prompting a try? Grab the 10-step chain prompt we
shared earlier for creating a blog post and replace the term [Topic] with any
topic that interests you. ChatGPT will assist you in crafting a well-organised,
SEO-friendly blog post in no time.
As you go through the chain prompts, be aware of the context limit. If
ChatGPT struggles to recall previous steps, think about how you can remind it
of the key elements needed for the next steps. You can do this by adding
relevant keywords, context, or main points from prior steps in your prompts,
making sure that the generated content remains consistent and focused.
By giving Chain Prompting a shot and adjusting the prompts to suit your needs,
you'll gain a deeper understanding of how this technique can maximise
ChatGPT's capabilities. Remember, practice is essential! The more you use Chain
Prompting, the better you'll become at guiding the AI through intricate tasks and
producing top-notch content tailored to your objectives.
You can find the 10 step chain prompt here to make copying and pasting
easier: https://www.chatgpttrainings.com/chain-prompt