In my last article, I lamented the explosion of ChatGPT prompts and their ineffectiveness for content creation.

I recommended that content creators forget “prompts” and instead think of “interviews and instructions.”

My most successful interactions with ChatGPT have been when I take time to prepare it and then converse with it, asking questions, probing for more information, and requesting that it dive deeper and deeper into an unwieldy sentence, unclear paragraph, or complex topic. I need to go that extra mile to get good results.

Through my explorations with ChatGPT over the last several months, I developed the five-step S.P.A.R.K. framework for getting the best content from ChatGPT.

S = Set up a creative brief

P = Prepare ChatGPT

A = Ask questions

R = Request a draft

K = Keep refining

By following this framework, you’ll be well on your way to getting the best content from ChatGPT and similar generative AI tools. By “best,” I mean usable, helpful content—the kind search engines respect, content creators delight in, and readers applaud.

Before we begin: What does Google say about using ChatGPT to create content?

In April 2022, even before OpenAI released ChatGPT to the public, Google’s stance on AI-generated content was negative. The search giant wasn’t happy with the idea of content creators filling the web with the kind of auto-generated content it considers spam.

But their position has since shifted. In November 2022, Danny Sullivan, Google’s public liaison for search, said that our focus should be on the content, not the tool we use to produce it.

In February 2023, Google further solidified its guidance. The headline? “Rewarding high-quality content, however it is produced.”

The post goes on to say that:

Using AI-generated content primarily to manipulate search rankings violates Google’s spam policies.
The appropriate use of AI or automation for content creation is not against Google’s guidelines.
As a content creator, your primary concern should be creating “original, high-quality, people-first content” that demonstrates E-E-A-T, or experience, expertise, authoritativeness, and trustworthiness. (More on E-E-A-T soon.)
If your readers might think, ‘How was this created?’ consider adding an AI disclosure to your piece.

Did I use AI to write this article? Yes—full disclosure—I used ChatGPT and other AI tools to help me:

Develop the hypothetical example we’ll soon dig into.
Research the issues around AI-generated content.
Create the S.P.A.R.K. acronym for my 5-step framework (the other options it gave me included B.U.I.L.D., S.C.O.R.E, and Q.U.E.S.T.).
Edit to improve readability.
Create a strong title for the article.
Identify possible gaps in the piece.

The framework is mine. The idea for this article was mine. The writing is mine. ChatGPT saved me a lot of time. And I hope that by following the S.P.A.R.K. framework that follows, you’ll have the same excellent experience.

Now, let’s look at each step of the S.P.A.R.K. framework through the lens of a hypothetical example. In it, I’ll collaborate with ChatGPT to create a blog post for a fictional company—a small, eco-friendly skincare brand called “NatureGlow.”

Step 1: Set up a creative brief—the foundation of content success

First, I created a brief with the following elements to let ChatGPT know the essential facets of the work to come:

Objective

The blog post aims to educate and inspire environmentally conscious consumers by discussing NatureGlow’s eco-friendly skincare products, emphasizing natural ingredients, responsible sourcing, and skin benefits.

Brand voice

Our brand voice is informal, empathetic, and passionate about nature; we’re a friend sharing the beauty of natural skincare.

Audience

Our audience is environmentally conscious consumers aged 25-45, interested in natural skincare products and committed to sustainability.

Content type and format

The content will be a blog post with informative and engaging narratives, visual content, and user-generated testimonials.

Key messages and themes

Messages include natural beauty, sustainability, skin wellness, connection to nature, responsible consumption, and community.

SEO and keywords

Use eco-friendly skincare, natural ingredients, sustainable beauty, and NatureGlow in the headings, meta description, and image alt text. Also, link to product pages.

Call to action

Encourage readers to explore the product range.

Legal and compliance

Adhere to regulations related to product claims, ingredient disclosure, and customer testimonials. Also, consider YMYL rules. [A note for the reader: YMYL stands for Your Money, Your Life. It’s a term Google coined to describe content that can affect someone’s health, happiness, safety, or financial stability.]

Examples and references

Cite authoritative sources for statistics, ingredient benefits, and sustainability practices.

Miscellaneous

Include content from influencers and eco-conscious community leaders. Highlight our partnerships and our commitment to reducing packaging waste.

This brief encapsulated the essence of the NatureGlow brand and gave ChatGPT a comprehensive roadmap for a blog post that resonates with the audience’s values and interests.

Step 2. Prep ChatGPT—add or update custom instructions and feed in your creative brief

With the creative brief under wraps, you’re ready to open ChatGPT.

If you’ve already added custom instructions, now is an excellent time to check the information to see if you’d like to make changes.

I checked my custom instructions and did not need changes in preparation for instructing ChatGPT to create the hypothetical blog post.

If you haven’t entered your custom instructions, now is the time.

Click the three dots next to your name in the lower left to access custom instructions.

There, you can tell ChatGPT about yourself and your business and how you’d like to receive responses, for instance, in tables or with real-life examples.

When you’re ready, feed ChatGPT your creative brief and tell the tool that you’ll now ask it questions in preparation for drafting the content.

Step 3: Ask questions—the art of broad-to-specific explorations

Before you ask ChatGPT to outline or draft your content, question and interview the tool, drilling down into the details that matter most. This iterative interview process is critical to engaging with an AI tool if you want to generate usable content.

For your first queries, go broad

Starting with broad questions will set the stage for your content and ensure it’s headed in the right direction. It will also pre-load your ChatGPT conversation with the details you want to include in the content.

For my hypothetical example, I asked ChatGPT this question: “Tell me about three popular natural ingredients used in skincare and why they’re beneficial.” No matter what ingredients ChatGPT came up with, I planned to assume they were in NatureGlow’s products.

For later queries, go deep, focusing on your brand

Next, focus the content more specifically on your brand and its products, services, and unique selling propositions. This step ensures ChatGPT tailors your content to aspects of your brand.

My next question for the hypothetical example:

“Can you relate those ingredients to the NatureGlow brand and how we responsibly source them?”

Continue to narrow down until you’re satisfied

At this point, continue your deep dive to eliminate ambiguities, answer potential reader questions and objections, and further direct the model.

Although I didn’t dive deep for the sake of our example, here are some questions I might have followed up with in a non-hypothetical scenario:

What artificial additives are in other natural skincare products, and are they harmful?
What does it mean when a farm is “certified organic?”
What are the top sustainable agricultural practices certified organic farms follow?
What synthetic pesticides might wind up in other skincare products?
What are the eco-friendly distillation methods you mention?
Can you find any statistics or research studies discussing quantitative details about minimized waste and conserved energy?

[Note: ChatGPT can access and provide links to statistics, research studies, and other information sources through plugins. I’ll be writing an article on ChatGPT plugins for content marketers soon. Follow me to be notified when I publish new pieces.]

Those questions would lead to a narrative beyond skincare to a story about harmony with nature, community empowerment, and a conscious choice for a better world, encapsulated in the NatureGlow brand.

Step 4: Request a draft—let ChatGPT write for you

For my hypothetical post, I next asked ChatGPT to use our conversation—and its vast stores of knowledge—to prepare a draft. The result follows.

1/2 - The first draft of a blog post created by ChatGPT. It looks good on the surface, but needs lots of attention before it's ready for readers. RENAE GREGOIRE

2/2 - The first draft of a blog post created by ChatGPT. It looks good on the surface, but needs a lot of work before it's ready for readers. RENAE GREGOIRE

Creating the post seemed magical. But our work as ethical content creators is not done by a long shot.

Step 5: Keep refining—adding the human touch to AI-generated content

Although the post is well-written, I wouldn’t publish without fact-checking, significant editing, and further dialoguing with ChatGPT for several reasons.

1. AI-generated content done poorly is still considered spam

Google can detect AI-generated content. The last thing you want is for Google to assume your post is spam.

How do you prevent your content from being considered spam? Use Google’s E-E-A-T protocol to evaluate your content based on experience, expertise, authoritativeness, and trustworthiness.

Looking at our hypothetical blog post through the lens of Google’s E-E-A-T, I see problems. You can also spot problems in your content by working through the following questions for each E-E-A-T element.

Experience—Does the content include first-hand or life experiences about the topic? Can readers tell whether the author has used what they’re writing about? In the case of our NatureGlow blog post, the answer is no.
Expertise—Can readers tell whether the author has the necessary knowledge—a blend of real-world experience and formal training—to write the piece? In the case of the NatureGlow blog post, the answer is maybe. Perhaps NatureGlow has the expertise to write about its products. Perhaps the company’s founders are cosmetics or skincare experts. Because this is a fictional company, there’s no way to know.
Authoritativeness—Do others point to the content author as an authority on the topic? Does the author train, certify, or inform experts in their industry? For our NatureGlow post, again, the answer is maybe. Maybe the NatureGlow blog has backlinks and citations from trusted websites. Perhaps the products have garnered positive reviews or endorsements.
Trust—Can readers trust this content based on the website’s credibility? Is the content accurate? Is the author transparent and authentic? Does the content meet Google’s experience, expertise, and authoritativeness criteria? For NatureGlow, the answer at this stage of the draft is no.

According to Google’s documentation, trust is the most important of the E-E-A-T factors. It represents the intersection of experience, expertise, and authoritativeness, plus information about customer service for online stores and peer-reviewed publications for academic authors.

2. The post contains overly flowery language

Another reason why the NatureGlow post needs work is because of ChatGPT’s use of grandiose, poetic language, which might turn some readers off:

True beauty blossoms from nature’s embrace.
They’re like a soothing whisper of nature.
Our chamomile is as pure as your intentions.
A calming caress for sensitive souls.

Then again, perhaps flowery language is appropriate for the NatureGlow brand, although nothing in my custom instructions or brief would have led ChatGPT to assume so.

3. The biggest issue is a lack of specificity

Despite my attempt at prepping ChatGPT, notice how generic the first draft of the content is:

We ensure fair trade practices—what does that mean?
We partner with certified organic farms—what does that mean? And what farms?
We adhere to sustainable distillation methods—such as?
Our partnerships with indigenous communities—which ones?

Those generic statements illustrate why you still need human intervention when using AI tools to generate content. I never told ChatGPT that we partner with certified organic farms, nor did I say that we partner with indigenous communities. The AI hallucinated those “facts” all on its own.

As ChatGPT’s human counterpart, you’re responsible for catching issues like those and asking ChatGPT to dive deeper where details are sparse.

When I spot a lack of specificity in content, it’s a clue that more content is needed, whether in the confines of the current piece or multiple pieces.

For instance, all ChatGPT had to offer about aloe vera is that it’s moisture-rich and natural, without artificial additives. If I published the blog post as is, I’d lose the opportunity to expound.

At this point, I’d likely add many more details to the current post and then spin off three more blog posts, one on each ingredient in the brand’s products.

The collaborative future of content creation: Humans plus AI

The allure of ChatGPT prompts is understandable, but relying on them can lead to generic, ineffective content.

By understanding the limitations and challenges of ChatGPT prompts, viewing AI as a collaborative tool rather than a standalone creative “entity,” and following the S.P.A.R.K. framework for creating content, you’ll develop a more effective content creation process and better content—the kind Google and readers love.

Originally published on Forbes.com.

5 Steps to Unlocking ChatGPT’s Power for Content Creation

Before we begin: What does Google say about using ChatGPT to create content?