Teaching writers when and how to use generative AI to create helpful content

March 19, 2024March 20, 2024 keywordana

Dive into a framework for teaching writers to use AI tools effectively while preserving expertise, factual accuracy and audience needs.

On Nov. 30, 2022, OpenAI launched ChatGPT, a groundbreaking AI chatbot that can understand and generate human-like text. Less than three weeks later, I started teaching writers how to use generative AI tools like ChatGPT to create helpful content.

While this may seem hasty, given the uncertainties around this new technology, I have experience adapting to innovative technologies that transform the writing process.

In the 1970s, I wrote using typewriters as a reporter and editor. Then in August 1981, I was introduced to a word processor at my job at Wang Laboratories.

This hardware device with a QWERTY keyboard and intuitive editing functions completely changed how I approached writing longer pieces. Just as word processors replaced typewriters, I believe generative AI will similarly revolutionize content creation.

What makes content helpful?

Now, before showing writers in the United Arab Emirates or content teams in the United States how to use the latest generative AI tools, I often begin by asking them, “What makes content helpful?”

After listening intently to their initial thoughts, I ask them to read Danny Goodwin’s article, “What is helpful content, according to Google?” As he explained on Aug. 19, 2022, helpful content:

Is created for a specific audience: Do you have an existing or intended audience for your business or site that would find the content useful if they came directly to you?
Features expertise: Is this content written by an expert or enthusiast who demonstrably knows the topic well?
Is trustworthy and credible: Would you trust the information presented in this article?
Meets the want(s) or need(s) of the searcher: After reading your content, will someone leave feeling they’ve learned enough about a topic to help achieve their goal?

It’s worth noting that his article was written before ChatGPT was launched. However, it’s important to reveal the bell curve that helpful content will be graded on before showing writers or content teams how to crank out more than 400 words of unhelpful content in less than 20 seconds.

What are the capabilities and limitations of generative AI?

It’s also crucial to have a clear understanding of what generative AI models can and cannot do.

Writers should know that these models can generate coherent and fluent text, but they may not always produce factually accurate or high-quality content without human oversight and editing.

Heck, even the AI chatbots admit:

“ChatGPT can make mistakes. Consider checking important information.”
“Gemini may display inaccurate info, including about people, so double-check its responses.”
“Claude can make mistakes. Please double-check responses.”

In the United Arab Emirates, I’ve frequently asked students at the New Media Academy to use a couple of AI chatbots to create a content calendar for the next 12 months on behalf of Visit Dubai, the official tourism website for that city.

Why is this assignment likely to produce a learning moment?

Generative AI lacks common sense, emotional understanding and real-world experience. It can generate nonsensical or misleading content if not guided properly.

For example, generative AI models create new data based on previous patterns. But Ramadan moves forward by 10 or 11 days each year in a 33-year cycle. So, different AI chatbots will put Ramadan-related content into different months in the coming year, depending on when their generative AI models were trained in previous years.

In the U.S., I might ask participants in training workshops to create a “wicked smaht” editorial calendar for Meet Boston, the official guide to Beantown (The Hub, The Athens of America and The City on the Hill).

Why? Because “Townies” know you can’t “pahk the cah in Havahd Yahd.” Plus, Bostonians have a few local holidays that are “rippah” – including Evacuation Day, Patriots’ Day and Bunker Hill Day.

What would they learn from this exercise?

Well, Gemini is more proficient at generating text based on prompts, mimicking various writing styles and producing content with a Boston accent. Its 650-word draft included:

January: “Fahkin’ Cold & Flippin’ Awesome.”
February: “Love is in the Air (and the Chowda).”
March: “St. Paddy’s Day Shenanigans & Springtime Awakenings.”
April: “Marathon Mania & History Buff’s Paradise.”
May: “Bloomin’ Beautiful & College Town Charm.”
June: “Get Your Red Sox On & Summertime Fun.”
July: “Fourth of July Fireworks & Freedom for All.”
August: “Beach Days & Back to School Buzz.”
September: “Fall Foliage Frenzy & Arts & Culture Extravaganza.”
October: “Wicked Spooky Fun & Halloween Hijinks.”
November: “Thanksgiving Feasts & Holiday Cheer.”
December: “Winter Wonderland & New Year’s Eve Celebrations.”

Dig deeper: Generative AI to create content: To use or not to use it?

When should writers use generative AI?

At this point, students or participants are ready to leverage AI chatbots for specific tasks. For example, generative AI can be particularly useful for certain writing tasks, such as:

Brainstorming and ideation: AI chatbots can generate ideas, outlines, or alternative perspectives on a topic. This can help content creators overcome writer’s block and explore different angles for articles, blog posts, or stories.
Researching and gathering information: Generative AI can analyze large datasets and summarize complex information. This can identify relevant sources, suggest interesting topics and generate strategic insights, saving writers research time.
Expanding on ideas or creating rough drafts: AI chatbots can help writers create basic content structures, generate factual descriptions, produce content in multiple languages or write simple introductions and conclusions.
Content Improvement: Generative AI can generate summaries or abstracts, identify grammar errors, suggest alternative phrasings and check for plagiarism. Writers can use it as a tool to refine their writing, not to automate it.

So, I believe that no matter which generative AI tools you end up deciding to use, you should start by asking if you’re doing the right things before turning your attention to doing things right.

How should writers use generative AI?

When it comes to doing things right, generative AI should be viewed as a tool to assist writers, not a replacement for human creativity and expertise. Writers should:

Develop clear expectations: Discuss responsible generative AI use in their writing assignments. Outline acceptable applications and emphasize the importance of original thought, ethical sourcing and fact-checking.
Focus on prompt engineering: Craft clear, specific prompts that guide an AI chatbot’s output toward its desired outcome. The specificity of the prompt influences the quality and relevance of the generated content.
Use critical thinking and editing: Emphasize that AI-generated output needs critical analysis and editing. Writers should assess the factuality, relevance and coherence of generated content, integrating it thoughtfully into their writing.
Incorporate the human touch: Add personal insights, experiences, or anecdotes to the AI-generated content to make it more engaging and authentic.

Does this approach help writers create helpful, reliable information that benefits people?

On March 7, Sara Lebow, a senior newsletter analyst for Insider Intelligence, wrote an article titled “Google’s update targets spam, but marketers can still use AI to create content.” She reported:

“Google’s latest core update took direct aim at generative AI spam, so can marketers still use generative AI to create content? In short, yes.”

“The new core update focuses on decluttering spam from search. Google is targeting sites using generative AI to mass produce low-quality content. Marketers can still use ChatGPT to draft content, but shouldn’t publish the raw output from these tools, according to Lily Ray, vice president of SEO strategy and research at Amsive.”

Maintain transparency when using AI-generated content

I should disclose that I used OpenAI’s ChatGPT 3.5, Google’s Gemini (formerly Bard) and Anthropic’s Claude 3 to generate drafts, outlines and ideas for this article.

But it’s worth noting that these generative AI tools generated about 400 words of content on this topic, which were structured like outlines or listicles and between two-thirds and three-quarters of their suggestions were overlapping.

Plus, I edited, fact-checked and refined their content to ensure quality and accuracy. Then, I added relevant statistics, incorporated credible quotes and included citations from reliable sources.

In other words, I followed the sage advice of Henry David Thoreau and avoided becoming the tool of my tools.

A new BrightEdge generative AI capability

I’ve also started using BrightEdge’s Copilot for Content Advisor, launched on Dec. 6, 2023. BrightEdge said Copilot for Content Advisor provides “all the insights necessary to develop content that performs in SEO. This includes audience and comprehensive SEO insights, structured content suggestions and AI-generated first drafts ready for human refinement and collaboration.”

BrightEdge also reported that over 85% their early users had “reported a significant improvement in their content publishing productivity.” One of these early users is Chris Pareja, Senior Manager of Online Marketing at Topcon Positioning Systems, who said, “Copilot for Content Advisor uses AI to save me 6-8 hours of research per content page created. As importantly, it identifies blind spots I either forgot or didn’t think about when trying to meet my prospects’ needs and understanding of our offerings.”

When I first tested BrightEdge Copilot, I thought it significantly reduced the time I normally spend on keyword research and dramatically improved the content creation workflow by creating a content brief before generating a first draft of a blog post or article.

After one false start while I was learning to train the AI with optional Inputs, Copilot for Content Advisor delivered a content brief in about 60 seconds and then quickly produced a rough draft, setting the stage for the writers I’m training to add their creativity and brand messages.

Since then, I’ve taught a series of training workshops for the content team of a company that provides digital products and services for higher education. I’ve encouraged the marketing writers to use Copilot for Content Advisor to generate drafts for 75% of their articles. I’ve also told them to add relevant statistics, incorporate credible quotes and include citations from reliable sources.

In other words, I’m practicing what I preach. Although it’s too soon to report any results, I did experience déjà vu. I remembered the first time that I’d used a word processor. And I recalled that I’ve never used a typewriter again.

And after teaching writers when and how to use generative AI to create helpful content, I’m confident that I know what will happen next. There’s no going back.