It’s no secret that the use of AI to generate content has skyrocketed in the past year. The catalyst was GPT-3, an AI language model by the nonprofit OpenAI. Using GPT-3 as the foundation, dozens of new technology tools are now making it easy to access and use AI for content creation. And the impact will be massive. By 2023, Forrester predicts that 10% of all Fortune 500 enterprises will use AI to generate content.
What does that mean for the content marketers so many businesses currently employ to develop their brand assets, blog posts, social messages, and more?
Many creatives might have a knee-jerk reaction and say AI can’t bring the creativity, judgment, and problem-solving skills needed to create high-quality content. And that may be true… today.
But be honest: not everything you do as a content marketer requires creativity, judgment, or problem-solving. Some of it’s tactical and tedious, and you’d be happy if it took up less of your time. And here’s another truth: Sometimes, content people feel… stuck. Could AI help?
To find out, I signed up for a free trial of Jasper.ai, one of the more famous content generators spawned from GPT-3. I wanted to experiment and see if using AI could help me write this blog post faster and (subjectively) as well as I would have done on my own.
I had another motivation as well—to give myself a reality check. Because AI for content generation is here. Pretending it isn’t won’t halt its impact on content creators, myself included. I need to know what it can do (and what it can’t) to prepare.
Here’s what I found.
With AI, Your Outputs Are only as Good as Your Commands
Based on instructions from the platform, I wrote a short “command” asking the AI to: “Write a blog post about how businesses will use AI to generate content in 2023, including how human content creators will benefit and how they can start experimenting.”
Here’s what it gave me:
By 2023, AI will be used by businesses to generate content. This means that human content creators will benefit from AI because they will have more time to focus on creative tasks. Additionally, businesses can start experimenting with AI now to get a head start on this trend.
Not great, right? The text itself is fine in that it’s grammatically and syntactically correct. But it’s also passive and a bit stolid. I could have edited it to create an active voice and a more engaging tone, but it wouldn’t have been worth it—the generated text wasn’t much longer than my command had been.
Seeing as this was my first time, however, I chalked up my poor first attempt as user error and tried again. This time, I asked Jasper to: “Write an outline for a blog post on the advantages and disadvantages of using AI for content generation, with examples of businesses already doing it.”
Here’s what it wrote:
1. What is AI-generated content and how does it work
2. The benefits of using AI for content generation
3. The potential drawbacks of using AI for content generation
4. How to implement AI-generated content into your business strategy
5. Tips for ensuring the quality of your AI-generated content
6. Examples of businesses that are already using AI-generated content successfully
That seemed like a reasonable outline, if basic. I shortened it to fit within Jasper’s character limits for its commands and then asked it to compose the full blog using the outline. This time, I got a paragraph for each of the outline items.
Let’s look at the paragraph it produced on the benefits of AI content generation:
There are several benefits to businesses that choose to generate their content using AI. Firstly, it is much faster than traditional methods. This means that businesses can produce more content in a shorter time frame, which is beneficial for meeting deadlines or increasing output. Secondly, it can help to improve the quality of content as it can identify errors and correct them automatically. This is particularly useful for businesses that produce a large volume of content, such as news outlets or blogs. Finally, AI-generated content is often more accurate than human-generated content as it is not subject to the same biases and error-prone nature of humans.
Again, this text is fine if I just look at the issues of grammar and general meaning. I, personally, would never use the word “firstly” (is that even a word?), but that’s easy enough to fix.
Yet, it still falls short for me because grammar is only part of what I think about when I write. As important are the ideas: they make the difference between a strong piece and a weak one. From that perspective, this paragraph needs two improvements: the ideas that are generally right must be clarified and substantiated, and the ideas that are not quite right need to be corrected.
With that goal in mind, here’s my revision:
Businesses have plenty to gain from using AI to generate content, particularly speed and quality. In the realm of speed, AI can generate paragraphs or even pages of content much faster than human writers can. This allows businesses to produce more content in a shorter time frame, which helps them meet deadlines or increase output. AI that can identify spelling or grammar errors and correct them automatically can also improve quality—which is especially useful for businesses with large volumes of important content, such as media outlets, law firms, and financial institutions. But user beware: AI-generated content can espouse biased or even toxic ideas if it’s trained using a data source that includes that kind of content. Any business integrating it into its workflow should build in human oversight to ensure the content meets business guidelines.
Better? More accurate? I think so.
Even with a Clear Command (for a human), AI May Not Give You What You Want
Most of the paragraphs produced from the outline were like the example above: fine grammatically but thin or borderline inaccurate when it comes to the ideas. I wanted to see if I could get the AI to write a paragraph with evidence if I asked it to and gave it sources.
My command read: “Write a paragraph using data from McKinsey, Gartner, and Forrester on the growth expected in the business use of AI for content generation.”
What I got was… well, take a look at the beginning and the end (separated by ellipses) of what the AI produced:
Accordingsey,%23.”.,-generatedating,writing.,’s,.-generated.,,. generation content for AI using of benefits many the enjoy can businesses execution and planning careful with … This 20 in 300 by grow to expected is generation language for AI of use the McKin to
The full, generated text is about 4-times longer and more of the same gobbledygook. At least in this case, for the question I asked and how I asked it, the AI couldn’t help me substantiate an argument with third-party sources. Another way of saying that is:
AI content generators don’t know if what it wrote is true. Humans, for now, still need to do research and validate the ideas in AI content to ensure accuracy.
The Bottom Line: I Need to Learn More
I could share more of the paragraphs I received from my experiment with content generation AI, but I think the takeaway is clear.
AI is already capable of producing content that resembles what a human writer does when producing a new blog post. It is not an out-of-the-box output, however. It requires editing—maybe even substantive editing. Still, an AI’s “first draft” could boost productivity.
As for me, while I can’t say I saved much time using AI to help me write this blog post, I think I’ll keep trying—because I have to. More practice will probably help me get better at writing commands in a way that gets the generated content closer to what I am looking for.
But equally important is the fact that AI for content generation is a real tool in the content marketing ecosystem now. If I want to stay relevant and ensure I develop new skills that only humans can, I need to know what AI can do—and what it can’t.
Stay informed! Subscribe to The Content Strategist for more insight on the latest news in digital transformation, content marketing strategy, and rising tech trends.