AI–artificial intelligence–specifically Generative AI (GenAI), offers the promise of nearly instantaneous content for our websites, blogs, email newsletters, social media posts, and just about anything else where our words are needed.
But is this what our audience wants? Fast, disposable content to dump in the landfill of our newsfeeds and inboxes? More generic content in a sea of vanilla?
Midjourney: /imagine: a close up of vanilla frosting shaped to look like a calm ocean with a model schooner on it zoom in far enough so you can't see the edges of the frosting [sic] painted –ar 16:9 –v 5
Or do they want to be moved? Seen and heard? Learn from our personal experiences? Consider our new ideas in our thought leadership pieces?
And is there a middle ground? Is there a way in which we can take advantage of some of the benefits of AI without losing our brand voice, and more importantly, our own?
There's a time and place for “commodity content”…content that's not meant to inspire or educate, to challenge or coerce. I don't need Bill Shakespeare writing my table saw manual or Maya Angelou explaining why my insurance won't cover my latest eye doctor visit.
Since the public release of ChatGPT in late 2022, content creation has rapidly evolved from a mostly human endeavor to an AI assisted affair. But too many companies and writers believe that a simple prompt-and-click at an LLM (large learning model) like ChatGPT or Claude 2 will generate content that will rank well and move their audience to action.
Unfortunately, like a cake mix that only requires water, the final product is often lifeless, unmemorable, and just like everyone else's. Little chance anyone's going to go back for seconds.
Midjourney: /imagine: someone eating a piece of cake with a disappointed look on their face in the style of Norman Rockwell –ar 16:9 –v 5
This article explores how to keep your critical human voice–whether it's your brand voice or your personal voice–recognizable amidst the rise of GenAI, providing valuable, unique insights that will keep your audience coming back for more.
The Risks of Over Reliance on Generative AI
Let's start by talking about some of the obvious problems when we lean too heavily on AI tools:
Generic copy. There are many safety features built into popular LLMs, which is a good thing. However, it robs the copy of a Point of View that can grab and engage your audience.
Lack of personal experiences to draw upon. While GenAI can pull from databases and libraries of other people's experiences and reflections, it has none of its own. There's nothing new under the sun.
AI can generate content for you, but only your humanity can infuse it with soul.
Lack of truly new ideas. GenAI models don't “think” the way we do. They take guesses at the most likely next word or words and go with that. While it can mimic human writing–sometimes with amazing results–it's not creating something truly new as much as it’s following an algorithm.
Hallucinations. Fake it until you make it. There have been plenty of examples of LLMs providing answers that are objectively false. The official term for this is “hallucinations.” The unofficial term for this is “making shit up.” If you're relying on GenAI to create an important document and you're not checking the facts you may be dooming yourself, your business, or steering your audience astray.
Biases. Also known as “garbage in, garbage out.” GenAI is trained on reams of data, often scraped from the internet. Could you imagine what would happen to a human who got all of their information unfiltered from the internet?!? <shudders>
There are plenty of examples of gender bias, racial or ethnic bias, cultural bias, age bias, you name it, in LLMs. Without human intervention your content can come across at best as tone deaf, and at worst as racist, homophobic, antisemitic, and so on. Not a good look.
Cutoff dates. Current LLMs have a cutoff date: the date at which no new information was entered. New events are unknown. For example, at the time of this writing, ChatGPT's cutoff date was September 2021.
Anything that happened after that date won't be in its database, so it doesn't know that Christine McVeigh died (coincidentally on the day ChatGPT 3.5 was released), or that Russia invaded Ukraine, or that Will Smith won Best Actor and earned a 10-year ban from the Academy Awards on the same day.
Copyright issues. It's hard to know how this will play out, and different countries are writing different laws around AI-generated content and copyright. Between the first and second drafts of this article, the US District Court ruled that AI-related content won’t earn a copyright. Future court cases might refine or completely reverse this ruling.
And the flip side of this concerns: as LLMs gobble up words and sentences written by others, will you accidentally be violating someone else's copyright?
Lack of human connection.
The point of content is to generate leads, right? No. The point of content is to create a connection.
With an audience? No. With a specific person. One specific buyer, prospect, individual who you have created a cozy space for in your mind's eye. That one person you think of when you sit down at your computer: fingers flexed, claws hovering over keyboard. "Dear Valued Customer..." you write. Then you BACKSPACE BACKSPACE BACKSPACE. "Hi, you," you write instead. "It's me."
Who you are, how you create a relevant connection with one person at a time is the key to marketing success in the age of AI. (BTW: It always has been... but it's never been more true now in a world where "CREATE MOAR CONTENT 20X FASTER!" is the promise of countless generative AI companies.)
Your From line matters more than your subject line.
How to Use AI in Creating Unique, Thoughtful Content
I'm not against using AI tools to help you create content. In fact, I've used AI to help me develop this article.
- I asked Claude 2 to help develop an outline based on my topic
- I shared the outline with ChatGPT and asked it to make improvements
- I shared that version with Claude 2 and asked if there were any further improvements or edits to make
- I asked both Claude 2 and ChatGPT for experts in the field that I could ask for quotes for the article
- I asked Claude 2 to review my rough draft and suggest improvements or omissions
- I used Midjourney to generate some images for this post
- I took my first draft and asked Claude 2 for any recommended improvements on tone, grammar, or opportunities for improvement
I ended up not using much of what the two LLMs provided for a framework, but it was a good jumping off point nonetheless.
Other content providers spend critical time upfront through “chat priming” and other techniques.
We harness AI technology to enhance content writing efficiency and performance while maintaining our distinctive brand voice. This is achieved through chat priming, where we input information into the AI upfront to ensure the output aligns with our brand identity. We compile a comprehensive "super document" containing our About Us page, communication style, top-performing blogs, testimonials, case studies, approved statistics, and other vital brand components.
Uploading this Super Document via Anthropic's Claude yields superior outcomes by providing contextual information prior to the writing process.
Here are some ways in which you can use AI to improve an article rather than trying to outsource it completely.
Ideation. Probably one of my favorite ways to use LLMs these days. If you've ever suffered from writer's block, then you know the pain of staring at a blank page. (And if you've never suffered from it, how dare you call yourself a writer!)
Of course, these initial prompts need more refining, some keyword research, and some human intervention before they see the light of day. All we're looking for is to move away from the blank page and into the light.
Research. Years ago someone told me, “Wikipedia is a great place to start your research, but a terrible place to finish it.” In other words, you can't always trust the information you find there. Well, LLMs are like that but turned to eleven. Not only are LLMs prone to hallucinations, most don't cite their sources.
You can avoid some of this by using the AI tools built into search engines, like BingGPT or Google's SGE (Search Generative Experience), which not only cite their resources but link to them.
Search Engine Optimization. SEO is critical to almost every business; after all, it helps put us in front of our prospects when they express what they need or what their problems are. While AI's impact on SEO and search results is worthy of its own post, here let's just talk about how AI can help us create content that more closely matches the needs of our ideal customers.
SEO tool companies like Moz and Ahrefs are building AI into their offerings. I've also been using MarketMuse, an AI-powered tool that can provide research and optimization services around topics or to improve already existing pages.
In this example, I want to see how I can improve the ranking of a page about basement waterproofing. MarketMuse looks at the top 20 results, sees what phrases these pages rank for, and then provides feedback on which keywords to include and how often to give you a better chance to rank well.
It provides a content score as well: in this case a 24, which is better than the average of 17, but below the target of 36.
It also provides a word count with some context: 714 words, while the average top 20 page includes 1,049 words and a target of 1,161.
This of course isn't everything you'll need to rank well; page and domain rank come into play, as do other factors. Still, if you're looking to make sure that your content matches up with the needs–and searches–of your ideal customer, AI can help.
Pro tip: you could copy and paste this content into an LLM and ask it for help improving the page!
Frameworks. Even if you plan on writing your post your way, having a framework of the most important topics and subtopics can help. GenAI may suggest an idea you hadn't thought of covering. It may simplify your writing process, as you only need to “fill in the blanks” under the bullet points it creates.
Grammar and tone checks. Even us English majors can struggle with certain parts of the language. (Is it “even us” or “even we?” Not sure.) And we've all been known to write in run-on sentences, use dangling participles, or just get the tone wrong. Running a near-finished draft through AI can iron out some of those issues.
These are some excellent suggestions, by the way!
How to Stay Human and Create Remarkable Content
Now that we've looked at some of the ways that AI can assist our writing, let's talk about how we can maintain our unique, human voice throughout the process.
Use AI as a helper, not as a replacement. I've hinted at this all along. I don't feel it's wrong or cheating to use AI in creating content, any more than it's cheating to use Google or “phone a friend.” You need to lead when it comes to strategy, direction, and tone.
Create an iterative feedback loop between AI and yourself. In the examples above you can see how I used AI to further the process, then took the reins, then got feedback, and continually iterated for the best piece of content I could produce.
When prompting AI for something detailed like a plan, give it lots of context from various perspectives. Then, before requesting the first draft, ask it to ask you more questions that will help AI develop the final output. Then do a final brain dump. You’ll be amazed at how your responses can help AI get to a much better result!
Don't apologize for using AI tools. Years ago–like in the early 2000's–I had a PITA client who told me he didn't feel he should have to pay the quoted price for the website I built. Calmly, I asked why.
“Because I've discovered you used a piece of software to help you build the website.” (It was DreamWeaver, for those of you old enough to remember.)
I thought for a moment, then asked, “if you hired a home builder and it turned out he used power tools rather than hand tools to build your house, would you ask for a discount? Or would you pay him more for the investment he had to make in those tools, the time it took to master them, and the consistency of all the cuts he made?”
He paid in full. Then I fired him.
The point is, power tools are not bad. We just need to be aware that they require a little more safety and precaution to keep everyone safe.
Tell personal stories that connect to the topic and your audience. See above.
Find a unique topic or angle. Remember, AI can't really create anything new. It's just remixing old songs.
Talk about the things no one else will talk about. Most people (AI or not) talk about the same things. You need to be the one to talk about the pain point of your audience.
Include quotes from industry experts. See above and below. Bringing more voices and more humans into your piece should enhance its ability to convince and convert.
Include relevant media from other sources. It's not that you can't ask AI for help or suggestions here, but if you're trying to create the most valuable piece of content possible, consider bringing in video, charts, audio, or any other element that will help tell your story or get your point across in a way that words cannot.
Be edgy or contrarian. While you can ask AI to write in a certain tone, there's only so far you can push it. While this may not be sound advice if you're writing for a CPA or law firm, see how human you can make your content with honest emotions.
Be funny. While AI can tell a joke, it can't make a joke. It doesn't know what funny is.
Don't believe me? Just ask it to rewrite a piece of copy and make it “funny.” It might throw in a dad joke or two, but no one's going to LOL.
Midjourney: /imagine: a robot with a microphone doing a standup routine at a comedy club drawn by sci-fi artist Jim Burns –ar 16:9 –v 5
Read it out loud. This is a tactic that I've been using since at least college. I want my written voice to sound like my spoken voice. This is sound advice even if you're not using AI as part of your content creation process.
When you think you have your final draft, read it out loud. What sounds awkward? Stilted? Inhuman? Fix those bits.
Conclusion
As I write this, I recognize that AI is constantly evolving. When I attended MAICON (Marketing AI Conference) this summer, more than once a speaker referenced that these “amazing” tools we are using right now will seem “quaint” just a year or even a few months from now.
In six months, AI may have gotten a sense of humor. In a year, hallucinations may be a thing of the past. In two years, AI may have gotten to the point of “human intelligence.” (Not an oxymoron.)
That doesn't change the main point of this article: that the best, most valuable content for humans will be written by humans. Humans who know how to leverage AI.
As Claude 2 proposed in its conclusion: “The future is a symphony of man and machine.”
Damn. I couldn't have put it better myself.
Rich Brooks is founder and president of flyte new media, a digital agency in Portland, Maine, that’s been in business for 25 years. He is a nationally recognized speaker on entrepreneurship, digital marketing, and social media.
He founded The Agents of Change, an annual conference and weekly podcast that focuses on search, social & mobile marketing. He recently co-founded Fast Forward Maine, a podcast and workshop series for growing Maine businesses.
Rich is the author of The Lead Machine: The Small Business Guide to Digital Marketing, a popular and well-received book that helps entrepreneurs and marketers reach more of their ideal customers online.