Prompt engineering: 10 tips for content editors using generative AI

Published on July 16, 2024

Prompt engineering: 10 tips for content editors using generative AI

We've all heard the term slop. Low-grade content produced by generative AI is splattered across websites, clogging up search results and adding volume to the internet without adding value. Their sentences go round and round in circles without getting to the point, and their fabrications, generalizations, and superficialities dilute the pool of knowledge we drink from.

But it doesn't have to be this way. GenAI can be a boon for content creators and editors when used correctly. By combining practical instructions or queries for GenAI—called prompt engineering—and leaving a human mark, content creators and editors have a reliable technique for generating high-quality output.

My name is Bulent, and I've been a content editor for a loooong time. I won't age myself by giving you an exact number, but my years of drafting, editing, publishing, and curating content have given me some acumen. I've also worked with various AI tools and have seen their potential to support the editorial process.

In this post, I'll share 10 valuable tips for prompt engineering. Together, we can liberate the world from banal content and turn the tide against slop. Ready? Let's go!

See also: 7 takeaways from our survey on generative AI professional usage

1. Be specific and clear

Clarity is vital for prompt engineering; vague instructions only lead to ambiguous results.

If you ask a GenAI tool to "Write about composable content," the answer will likely be generic and rambling. For a more impactful result, include the following in your prompt:

  • Desired word count.

  • Main points to cover.

  • Specific topics/subtopics.

  • Unique angles/perspectives.

Let's try an example input: "Write a 600-word article regarding the impact composable content has on the content management software market, emphasizing increased productivity and new collaboration tools." The output will be more attuned to your needs.

2. Use examples for style

An effective way to communicate what you're looking for is to share examples in your prompt. You can add copy snippets in your preferred style, plus examples of things to avoid. It's also helpful to describe the characteristics of articles you admire, showcasing specific techniques or elements you'd like to emulate.

For instance, let's follow on from your initial prompt about composable content: "Write in a style similar to an article in National Geographic, balancing scientific accuracy with engaging storytelling. Avoid overly technical jargon, but don't shy away from introducing and explaining complex concepts."

Use examples for style

Sample output from ChatGPT 4o

3. Provide context

Context is also essential when it comes to AI-generated content. The more background information you provide, the better AI can produce the desired output. Consider including:

  • Target audience: Are you writing for students or retirees? Musicians or lawyers? Content creators or front-end developers?

  • Content purpose: Is this an informative piece, a thought leadership article, or a how-to guide?

  • Publication medium: Is this for a blog, a professional journal, or a social media post?

  • Brand voice: Are there any specific tone or style guidelines to follow?

An example prompt based on the above is: "Create a blog post for CTOs at medium-to-large enterprises considering the implications of replatforming from a legacy CMS. The tone should be informative yet approachable, addressing common questions about migration, data security, and cost."

4. Break down complex tasks

When working on longer, more complex pieces—like white papers or case studies—it helps to break the task down into smaller chunks. This approach makes the process more organized and results in more coherent and detailed content.

Try asking the AI tool to generate an outline. For example, "Create a detailed outline for a 2000-word article on personalization and composable commerce, including an introduction, at least four main chapters, and a conclusion."

Once you have a suitable outline, you can work your way through each section step-by-step, like so:

  • "Expand on the introduction, providing a hook and briefly outlining the main points to be discussed."

  • "Develop the second chapter with a definition of personalization and some examples of personalization in action."

  • And so on.

This segmented approach allows for more control over structure and depth while affording you space to assess content quality as you go.

5. Experiment with different angles

Prompt engineering is equal parts art and science mixed into a heady brew. (Drink deep, friends!) If your first attempt at a prompt doesn't yield the desired results, feel free to scrap it and start over. Try rephrasing your prompt, adjusting its details, or approaching the topic differently. This spirit of experimentation can lead to exciting and unexpected results.

For example, a prompt like "Explain headless CMS" will probably come up short because you ignored our first tip about clarity. But there's no cause for alarm; you can try a variation like:

  • "Describe headless content management systems as if explaining it to a 12-year-old."

  • "Compare and contrast a headless CMS with traditional content management systems."

  • "Outline the five essential features of headless CMSes and their implications for content management."

If no single answer delivers the goods, you can patch together a sequence of answers to build the ideal response. Keep note of which prompts work well for different types of content. Over time, you'll develop a playbook of what's most effective.

Experiment with different angles

Sample output from Claude 3.5

6. Leverage AI capabilities

Something to be aware of is that modern AI tools can do much more than simple text generation. They offer a full range of capabilities that can inspire your creativity.

A great one for editors is brainstorming ideas for your content pipeline. An example prompt: "Give me 10 unique ideas for blog posts about JavaScript frameworks that haven't been widely covered." Some results will be terrible, but if two or three gems emerge for further development, you'll be glad you asked.

If you're working with a longer text and are pressed for time, you can use GenAI to provide a summary before you dive in with the blue pencil. This can significantly speed up your editing process, allowing you to focus on the most salient aspects of the content.

Also helpful is fact-checking. Let me put this here in BIG RED LIGHTS. 🚨 An AI itself should not be relied upon to perform actual fact-checking. 🚨

However, it can prove its worth by highlighting statements that seem to make factual claims, suggesting areas that might benefit from verification, and generating questions about specific points that could guide human fact-checking efforts. This underscores the importance of your role in the content creation process.

You can also use GenAI to bring in multiple perspectives, e.g., "provide three different counterarguments to the main thesis of this article," or roll the dice on language translation and localization, e.g., "adapt this content for a German-speaking audience, considering cultural nuances."

7. Iterate and refine

The most sensible approach to using GenAI is to view its outputs strictly as a starting point, not a finished product. Bring your expertise as an editor to refine the generated content's language, structure, and flow. Add human touches, personal anecdotes, or industry-specific insights the AI won't have.

You can work with a GenAI tool in multiple stages of the production process: first, by generating the initial draft, then by asking for improvements and expansions on specific sections, and finally, by requesting a final polish that addresses tone and style consistency.

The goal is to combine AI's efficiency with human editing's nuance and creativity. Some folks refer to this work mode as becoming like a "content cyborg." In less excitable terms, I think of it as having a virtual writing partner, and we can bounce ideas around.

8. Consider ethical implications

This is the big one. As GenAI becomes increasingly prevalent in content creation, it's the editor's role to safeguard and maintain ethical standards. Obviously, we must verify any facts and statistics provided by the tool.

A more subtle thing to be aware of is the potential biases that can emerge in AI-generated content. For example, when asking for a story about a doctor and a nurse, you might find the output defaults to presenting the doctor as male and the nurse as female.

The Berkeley Haas Center for Equity, Gender and Leadership studied 133 AI systems across multiple industries and found approximately 44% exhibited gender bias, and 25% showed both gender and racial bias. (H/T UN Women)

As a final measure of protection, follow your organization's guidelines on AI usage and disclosure, stay informed about copyright and intellectual property issues related to AI-generated content, and always run your text through a plagiarism detection tool.

Consider ethical implications

Sample output from ChatGPT 3.5

9. Optimize for SEO

Creating content optimized for search engines is mandatory for modern digital publishing. While it's no substitute for working with skilled SEO practitioners (or for reading the Contentful SEO guide), you can use GenAI to incorporate a few SEO best practices into your prompts.

If you have a target keyword, include it in your initial prompt. You can also ask the GenAI tool to generate SEO-friendly titles and meta descriptions or request that the content be structured for featured snippets with clear headings and concise answers.

For example, "Write a 1000-word article explaining the concept of structured content and its benefits, optimizing for the keyword 'structured content.' Include a FAQ section suitable for a featured snippet."

10. Use persona-based prompts

Another technique you can try is experimenting with different voices or perspectives to add depth and variety to your content. Try prompts like:

  • "Explain content design as if you were Sarah Winters addressing a group of government and public sector workers."

  • "Write about the future of storytelling from the perspective of a technology evangelist with 20 years of experience."

  • "Describe the process of building a basic ecommerce site in the style of a community developer advocate, sharing tips with beginners."

This approach can help create content that resonates with specific audience segments or adds a unique flavor to your pieces.

Fighting the good fight against slop

There you have it. Prompt engineering is a skill that can only get better with practice and experimentation. As you work more with GenAI tools, you'll develop an intuition for crafting prompts that yield the best results for your specific requirements. 

Using the tips outlined above, you should be better equipped to navigate the exciting intersection of AI and content creation. Best of all, you won't contribute to the existing mountain of slop. 

Remember to stay curious, refine your techniques, and balance AI efficiency with human creativity and oversight. Thank you for reading!

See also: 7 takeaways from our survey on generative AI professional usage

Subscribe for updates

Build better digital experiences with Contentful updates direct to your inbox.

Related articles

An enterprise CMS is a content management system that meets the needs of enterprise-class companies. Read on and decide if your organization needs one.
Guides

What is an enterprise CMS?

January 19, 2024

Static and dynamic websites each have their own advantages. This post explains the practical differences for developers and helps you choose which to use.
Guides

Static vs. dynamic websites: Building the best website for your users

April 2, 2024

GraphQL variables let you create reusable queries and mutations that are type-safe. This article explains how to use them, with examples.
Guides

How to use GraphQL variables to give queries type safety

October 22, 2021

Contentful Logo 2.5 Dark

Ready to start building?

Put everything you learned into action. Create and publish your content with Contentful — no credit card required.

Get started