- To get the most from AI content generation, you need to understand its limits
- AI content generation is like an enthusiastic new employee, not an experienced pro
- AI-generated content tends to communicate features, not benefits
- It’s not about AI vs human; it’s about AI and humans collaborating
- Marketing has always been about selling the sizzle, not the steak
Remember the last time you hired an intern? Maybe a recent graduate with a degree in your field; maybe a young school-leaver who wasn’t academic but had a capable head on her shoulders. They didn’t know your inhouse processes yet; perhaps they weren’t sure how office life works at all. But they were keen. And that’s what you saw in them. Potential.
2023’s hot topic – AI content generation – is a lot like that.
The mistake many companies make is thinking tools like ChatGPT and Jasper.ai are experienced and qualified professionals who can work unsupervised. They’re not. But with the right guidance, they make terrific interns. Energetic, ready to learn, and grateful for the opportunity. And unlike real interns, they won’t complain about working for peanuts. (Yet.)
In this blog, we’ll compare the pros and cons of AI content generation. How it best fits into your content marketing strategy, and why treating these tools as new trainees rather than experienced managers is the best way to get results.
How AI-generated content works
First, it’s useful to know, how does AI create content? Let’s emphasize that artificial intelligence is not “intelligence” as we know it. Just like the predictive text that guesses your words when you send an SMS, language models like GPT-3 look at the input text (using advanced statistics and complex calculations) to predict which sequence of characters is most probable to follow. Based on the millions of examples, it was trained on.
If that AI has been “trained” on a large number of examples, the resulting predictions can seem very convincing (The GPT3 model, which many AI tools use, is the result of 175bn parameters). This makes an AI writer truly excellent for producing lists of selling points, descriptions of products, the main ideas of a topic, and other “hard” outputs.
AI content generation, however, is less good at what readers find engaging and interesting: providing a reason to read on, with the welcoming warmth of a human personality. As you’d expect. It’s not human.
Tips for working with an AI writer
So what’s the best way to work with one of these tools? It’s to imagine they’re fresh to the job, just like an intern.
Think of what you do when a young new starter arrives in your team. (Assuming you’re a kind person.) You don’t put them to work on a €1m project straight away. You introduce your team, demonstrate how processes work, assign them a desk and make sure they know where the Fire Exits are. You provide detailed information that makes them comfortable.
The more precise and actionable your information is, the better that intern will perform. Because you’ve set the scene and given examples. What’s more likely to get a good result: telling an intern to “Organize the filing system” … or stating “Please remove all files older than five years and arrange them in alphabetical order?”
Despite that word “intelligence”, AI content creation works best when you assume as little as possible. Treat AI tools as junior assistants, not as members of the C-Suite. With that agreed, what are their advantages and disadvantages?
Ai content generation: the pros and cons
As the next industrial revolution (possibly) AI text carries the same pluses as previous mechanical innovations: do more, do it faster, and do it cheaper. AI content generation can produce thousands of words a minute, versus a few thousand a day for a human writer; it can also slice and dice that content into many variations for different audiences, letting you scale up your web footprint fast.
It’s also low-cost: even the poorest denizens of Upwork and Fiverr charge a few bucks a piece, whereas AI content creation tools are often free. AI tools can also SEO-ise your web pages at scale, thanks to its vast database of text snippets and the relationships between them.
There’s a downside. Obviously. Without human morals, AI content generators have no qualms about lying through their virtual teeth: outrageous untruths are common. And since they’ve been trained on a snapshot of the internet, they have all the internet’s biases and misconceptions within them – some of which aren’t nice.
The biggest issue of all, though, is that these tools aren’t “creative” in the human sense. They don’t form concepts or engage in abstract thought; they’re just running probability patterns on a vast quantity of data. And since many of today’s AI tools use the same underlying database (OpenAI’s GPT3) the output they produce can look very, very samey.
Let’s see how that works in practice.
Comparing AI-generated content to human creativity
Here’s a ChatGPT response to “sell me a muffin”
Welcome to Muffin World! We have the best muffins in town! Our muffins are freshly baked every day with the highest quality ingredients. We use real butter, fruit, and sugar to ensure a delicious and flavorful muffin. Our muffins are soft and fluffy, perfectly sized for snacking. Plus, our muffins are super affordable – you can get one for just $1.50! Stop by today and grab a muffin for yourself – you won’t regret it! Our muffins are guaranteed to brighten up your day.
Now as a description of the selling points of a muffin, it’s just fine. But read it again. You’re bored already, aren’t you. Because it sounds like something any muffin shop in the world would write. That alone disqualifies it from go-live status. Ultimately, the outcomes of AI tools are averages – and that’s the opposite of a strong brand with a distinctive tone of voice.
We gave a human writer the same prompt
No breakfast. No time. The Zoom’s done but now it’s 9.30 and you’re due downtown. Your hands feel jittery, your head’s starting to spin, you can’t think straight. Maybe your next meeting – and it’s a biggie – won’t go so well. Relax. At the Muffin World drive-thru, there’s a tray of blueberry-jammed deliciousness just out of the oven, handmade with the creamiest butter and juiciest fruit from local hipster farmers with tattoos and goatees. And if you drop by before 10am they’re just $1.50. (The muffins, not the farmers.) When you’ve got no time, it’s time for Muffin World.
Very different, isn’t it? It has a distinct voice. It uses emotion, creating a picture in our minds of that time we craved something sweet. There’s even a bit of humour, finishing with an offer that precisely matches the need. It’s an appeal to the heart, not the head – which for most products and services is how customers make their buying decisions.
And that’s the real difference between AI content generation and human talent. Not what they say … but how they say it.
Ways to use AI content generation
By now you’ll have an idea of the best ways to use these new technologies. Ask them to rough out a structure for a white paper. Ask them for 1,000 variants of a call-to-action message. Ask them about SEO keywords, the laws of physics, and how things work. Just don’t ask them to write your final content. Because if you do, it’ll lack most of the things that make you “you”.
AI tools aren’t even equivalent to our friends, much less our bosses. They’re helpful assistants, willing researchers, and hard-working partners for repetitive, high-volume, or just plain boring tasks that need to happen at scale. And when you use them in the right way, they can be amazing assets for your business.
AI: your new favourite intern
That’s how to use AI content generation tools: as that intern that’s just joined the team and is willing to do anything without complaint. Best of all, they don’t need to go home to sleep.
They can’t do everything. But if you understand what they do best, they can be very, very productive.
We’re using them at Contentoo to scale up content marketing for our clients – Check out our AI tool, you can sign up for our Beta program here.