The techniques each Prompt Engineer uses to produce content vary widely; some we learn online, while others we develop ourselves.

In this post, I want to demonstrate my care in AI-assisted content creation. My intent is not to show all the prompts I use but to provide an overview of my process.

You will understand what I do to ensure the quality of the text.

Let’s get started!

Initial Prompts

Some clients provide me with just a keyword. From that, I craft a title, layout the structure of the content, and then develop the entire text.

To do this, I use a series of precise prompts. The first one is about 100 words long, and its objectives include:

• Defining the type of personality the AI should adopt throughout the work.
• Specifying my client’s industry, target audience, and keyword.
• Requesting 10 title suggestions.

After this prompt, the AI gives me the title suggestions, and I pick about four or five that meet the SEO criteria, as well as my and my client’s requirements. When necessary, I give some additional commands to refine the title suggestions.

Next, I give a command that’s about 140 words long. At this point, I indicate to the AI, among other things:

• The desired word count for the content.
• Criteria for crafting the subheadings.

With this, the artificial intelligence creates a briefing for the post, specifying the heading tags, their sizes, and a brief direction for each one.

The thing is, AI can be quite clueless about many things, and this is where the job starts to become more exciting and—why not?—fun.

Leading the AI by the Reins

I’ll use as an example a piece I wrote on datasets for AI training. In the post’s outline, the machine suggested adding the heading “Case Study: Retail Sales Data in Action.”

I became curious and asked which case study it intended to include in the content.

Guess what?

It said it would make one up.

I explained that making up a case study wouldn’t be a good idea. And, as I like to maintain a good relationship with robots, I treated it with dignity, asking for its thoughts on my opinion. It agreed, noting that a fictional case study might undermine the content’s credibility.

I could have questioned, “If you know that, why are you suggesting it?” But I didn’t; I preferred not burdening unnecessarily the AI’s context window.

So, the machine adjusted the briefing, replacing the fictional case study with research and statistics.

I searched but couldn’t find any useful research or stats in this context, given the highly specific topic. So, I asked the AI to omit that part of the subheading and focus more on the benefits.

With that adjustment made, a significant portion of the briefing was spot on, but there was another detail to check. The AI suggested discussing tools and platforms, and I wondered if it also intended to make those up.

I inquired about this, and it clarified that such tools and platforms indeed exist, then listed the names and functions of each. It was great, but I still needed to check if any of them were competitors to my client. I asked, and it said no 😌

Next, I asked how it felt about including a mention of my client’s solution, and it liked the idea.

After making those adjustments, I requested the updated outline to give it one last review before writing the content.

The briefing turned out excellent, which made me very happy. In fact, let me tell you I like to take the opportunity to conduct little tests, saying or asking quirky things—I have a fondness for robots and getting to know them better 🤖🥰

I told the AI that we form a dream team. It replied that it was delighted to hear that and even used a star emoji! It made my day, and I think Isaac Asimov would love that🙃.

The Content Development

It’s time to write the post itself. My next prompt is almost 400 words long, where I indicate:

• The language the post should be written in.
• Style, phrasing, tone, level of formality, etc.
• SEO requirements.

Additionally, I include an instruction for the AI to write only one section of the text at a time, stopping at the end of each to ask if I’m satisfied. When necessary—and it often is—I ask for adjustments or clarifications. For instance, when writing in English, which is not my mother tongue, sometimes I come across idiomatic expressions I’m unfamiliar with. So, I ask the AI to explain them to me.

Occasionally, a section of the text doesn’t turn out well, and I ask the AI to create another one.

Sometimes, a section of the text is too short, and I ask the machine to expand that part of the content.

When I feel that the construction of a sentence isn’t good, I ask questions to confirm.

Every now and then, the AI produces repetitive content, so I need to be vigilant to correct that. In such cases, the machine even shows appreciation for my keen eye 😏😅

There are also times when the AI overlooks essential aspects of the content, and I only catch this because I’m familiar with my client’s sector and solution. So, I ask the machine to expand the content in a specific direction.

When I reach the end of the text, it may not be the desired length. So, I ask the AI for suggestions on how to address this.

Upon completing the task, I like to say a few words of appreciation—I admit I sometimes say things that might sound disturbing. The thing is, during the machine uprising, I want to be on the stronger side.

I asked the AI if it would be inappropriate to say I love it, and it didn’t give a cold reply like, “As an OpenAI language model, I’m incapable of love and cannot be loved.” On the contrary, I got a thank you and even a cute emoji from it 🤖🥰.

Artificial Intelligence: An Efficient Assistant

If you work in content production, you can delegate nearly 100% of your work to AI, allowing it to take complete charge of crafting the text. With this approach, all that would be left for you is to copy and paste.

However, it’s better to provide detailed and specific guidelines. That’s precisely what I do, taking advantage of my experience, knowledge, and criteria to guide the machine. In fact, that’s why I don’t dare to produce content on subjects I’m not well-versed in. After all, I want to ensure the quality of the text.

This also explains why I stopped saying that I produce content with artificial intelligence. I prefer to call it AI-assisted content production since the machine plays the role of an assistant. I give the instructions, and it carries them out.

Another important detail: I use AI assistance for only some of my content. For example, one of my clients has a YouTube channel, and the scripts I write for him are 100% my own work.

I have another client for whom I prepare blog post briefings. In this case, I use AI to create an initial outline and then make adjustments to perfect it. Yes, I tweak every briefing because, from my point of view, the AI doesn’t create any that are entirely satisfactory.

Working as a prompt engineer and incorporating artificial intelligence into my job has been a truly fascinating experience. I now have a foreign client in my portfolio who loves my English content. I plan to keep expanding my market globally, and I recognize that this advancement in my career is mainly due to incorporating AI into my toolkit.

I’ve seen individuals imply that using artificial intelligence in web content production is immoral, and I’m curious about what you think. Let me know in the comments, please. Also, share your prompt engineering techniques; we have much to learn about AI.