I’m going to spoil the article for you right away. No serious business that relies on long-form writing as part of its marketing strategy should use AI to write the final product.
That was a bold statement. I’ll spend the rest of the article explaining what I mean by that. To be clear, I am not advocating against AI in the planning process, just for the writing work.
Here are the main points that I’ll cover
Let’s get started.
First things first, there has already been a major study on AI vs Human Writing, and the results were fairly clear. Humans tend to find that the AI-written copy is more convincing.
So, how do I square that with my anti-AI stance? Well, let’s look at the MIT study that everybody cites to prove that AI writing is better.
If you spend a bit of time skimming the study, you’ll notice that the copy they’re working with is less than 100 words long. If all it takes to convince your customers to buy your product is 100 words, then you probably have a short sales cycle. Or potentially, visual marketing pieces like product demonstrations might be more critical to your sales funnel. So, my point still stands for longer-form writing.
In most cases, particularly for B2B products, your prospects will interact with you over several touchpoints. For instance, they’ll read LinkedIn posts, case studies on your website, white papers, and of course, long-form blog articles. AI writing won’t replace that because it’s simply too repetitive. You’ll see what I mean by that very soon.
Let me tell you a bit about myself. When ChatGPT came out at the end of 2022, I immediately began playing around with it to see how effective it was at writing articles. I worked as a freelance content writer back then, and I wanted to expedite my tasks.
For the first few months, it was awful. That was until ChatGPT4 came out - a massive improvement over the previous versions. Hence, I started using it on my personal blog to pump out content faster.
The main problem was that I was writing a travel blog, the whole point of a travel blog is to share your personal experiences - that’s what people come for. At first, I thought the content was pretty well-written. In fact, I think this exact feature makes AI writing so tantalizing. The first time you read it, you think to yourself “Hey, this is pretty good.” It’s a damn sight better than the type of content you used to see spinning out from content farms. Nowadays, those content farms have replaced their cheap writers with even cheaper AI.
However, as you’ll see in the examples later, practically every AI-generated article follows an identical pattern unless you explicitly ask it not to. Even if you tell it to adopt a different tone, it usually sounds so unbelievably artificial and hokey that you might just vomit all over your keyboard after reading it enough times.
That’s an exaggeration. But after a few months of using ChatGPT4 on my travel blog, I soon grew tired of reading paragraphs like this:
Darling, what a mess!
I constantly had to coach ChatGPT not to use words like “darling” as you see here. Other weak points are “bustling” and “vibrant”. How bustling and vibrant can a coworking space be? I’m just trying to get my work done.
It was almost like trying to exterminate pests. Once I got rid of bustling, vibrant, and dynamic, “darling” appeared. Instructing ChatGPT not to overuse words in general doesn’t work; you must explicitly instruct it not to use certain words. Eventually, I developed a list of words that it was strictly prohibited from using. Here’s how I did it:
So, I took that list into a different chat thread and instructed ChatGPT to never use them. Every few prompts or so, it needed to be reminded of this fact, because ChatGPT has the memory of a goldfish, a goldfish that swallowed a thesaurus. You can email me if you’re interested in seeing the full list.
None of those words are bad on their own per se. I actually still permitted it to use “Historic” and “Lively”. The issue is that ChatGPT would constantly use words like bustling, quaint, etc. in place of more meaningful phrases or sentences. If you’re willing to read a full blog post about Madrid, you would be more interested in hearing how a particular neighbourhood made the author feel, in their own words than just hearing that it’s “quaint”.
And that’s the crux of the problem with AI; it glosses over important details while adding mountains of generic filler content that sounds smart while being irrelevant to the reader’s decision-making.
And that’s what we’re all about at Cerebrum: helping the reader make effective decisions.
Side note, I had to explicitly tell ChatGPT to discuss things like SIM cards and tap water in my outline. These are things that I knew would be immediately important to a reader given my own experience as a digital nomad. I also had to quickly find a Reddit thread about finding a SIM card in Spain and copy-pasted the SIM card recommendations. ChatGPT is not capable of research at such a granular level.
While I’ve come out pretty strongly against using AI writing, there are some situations where it would be more efficient.
Like I said before, if you’re just running a content farm where quality doesn’t matter to your target audience, AI is a whole lot cheaper. These niches are typically less competitive, and your core readers might not be fluent in English, so what you say doesn’t really matter. The same would also apply if you’re marketing to kids or your target audience is just really naive.
Another use case for AI writing is the one from the MIT study, I could at the very least see using AI for product descriptions. However, writing things like product descriptions and other quick tidbits of copy has never been my domain; I focus on long-form content.
When it comes to brainstorming angles for an article, AI is pretty good. Similarly, I can leverage AI to create things like templates in niches that I’m not as familiar with. I still have a paid ChatGPT subscription because it makes the research process so much faster for me than before.
In basically all other cases, I think human writing is far superior to AI writing, especially for long-form content. I know saying “always use human writing” doesn’t sound pretty helpful, but I think I made a clear case.
Every time you publish an article, you’re trying to put your best foot forward. If your business matters, then your words matter.
In the coming years, people will get better at spotting the typical patterns in AI writing. For instance, if every travel blog overuses words like “bustling”, “dynamic”, and so on, readers will clue into it. Someone searching for authentic advice about a destination will skip over your blog when they read yet another boilerplate intro.
I’ve been leaning heavily on the travel blog example, but you’ll see more soon.
If you’re writing for a B2B audience, for instance, your customers will require more depth than what the typical AI article provides. Remember, the average B2B sales cycle lasts around 84 days.
You can bet that during the sales cycle, they will engage with both you and your competitors several times. If you all offer the same uninspiring insights in each content piece, that won’t play in your favour. Even worse, if your competitors use AI tools, they’ll invariably all end up writing content with a similar tone, similar generic insights, and so on.
So, the answer is clear, if one of your competitors uses insightful human-generated content that features your product, then they’ll easily be able to stand out.
As a final note regarding human writing, I should emphasize that you can rank your content with both AI and human writing. Given an excellent backlink profile, you can rank just about any garbage.
The point of writing exceptional content is less to help you rank, though it can help. Instead, you should write content to engage your readers and nudge them towards a purchase. You should write content that educates them on how to solve their problems, and how your business can help.
At this point in time, AI is not sophisticated enough to do those things without extremely heavy-handed guidance and editing. It would be faster to simply hire a professional content writing agency like Cerebrum or even write the content yourself.
I think that in the coming years, having the awareness to detect whether content, both in the written format and others, is AI-generated will be an essential skill. There are already plenty of studies showing that experts can’t tell the difference between AI writing and human writing. However, given some of the patterns I’ve noticed with AI writing, I detect unedited, unfiltered AI content almost immediately.
To prove my point, I decided to ask a popular AI writing tool, known as “Rankwizard” to help me write an article on the keyword “Agile Remote Working”. For some reason, the AI tool decided to write the article from the point of view of a content creator. The text is fairly saccharine, you can read the whole thing in the Google Doc I linked.
Here are some key tells that this content is AI-written.
The introduction to the article itself wasn’t bad, just a bit vague and lacking a hook, but the article soon fell off with a phrase that I’ve become all too familiar with when leveraging AI tools for writing.
Did you know we live in a fast-paced and dynamic and ever-evolving and vibrant and bustling work environment? Jokes aside, the opening sentence sounds OK at first glance, but pretty much every AI-generated article opens with something like this. Sometimes they open multiple sections in the same article with this phrase. That’s what the Rankwizard AI article did:
A best practice for content writing is to begin articles with a hook. A hook can be a statistic, a controversial statement, a quote, or anything to make the reader want to read your article to the very end. It should be thought-provoking. You want to be a thought leader after all, don’t you?
Let’s take an example straight from the current top-ranking article for “agile remote working”.
In the introduction, McKinsey, the top management consulting firm in the world, offers several statistics to pique your interest. And by the way, I don’t think that their article was very engaging either. But it’s still miles ahead of the AI one.
How do I know McKinsey didn’t use AI? They wrote that article in 2020, and they can easily afford to hire competent writers, so even if you don’t trust me to detect AI content, it’s hard to argue that this article was AI-written. Who exactly went and gathered those statistics and crafted that infographic? In 2020, it was definitely a human.
I’ve talked a lot about how human-written content is insightful, so long as you’re working with an expert writer. Maybe you haven’t seen enough evidence of that yet.
Since the AI-written article in the Google Doc is so boring, I couldn’t even bring myself to read the whole thing. It’s 4600 words for Pete’s sake. Here is a specific section that demonstrates exactly where things go wrong.
None of the facts in this section is incorrect per se but the writing is quite monotonous and not particularly useful. It just says that virtual activities are really good, and you should really do them because you can boost team morale.
Does it explain how a particular business can apply virtual activities? Of course not. AI does not have that kind of foresight unless you guide it through every single paragraph, it's like teaching a baby how to walk.
Compare this to how the McKinsey article covers this topic.
That’s just so much better. McKinsey immediately offers a concrete example of how a remote team improved its cohesion with team-building activities. They could also have given some instructions for team-building games, but either style would have worked.
I should also note that so far, the McKinsey and the Rankwizard AI articles look suspiciously similar. I would not be surprised if, in the back-end, Rankwizard is merely telling the AI to rewrite the content from the top-ranking article in its own words, and disallowing it from copying its specific examples since they are too unique to each article.
By skimming the headings, which most readers do anyway instead of reading the article, you can tell that it’s a snoozer. It doesn’t take any risks, it doesn’t offer any real-life examples, it doesn’t make any bold claims, it doesn’t offer any relevant case studies, and so on. People like to read the headings and then pick out an interesting section that might be relevant to their situation. The AI article does not allow you to do that.
To add those things, you would likely need to bring in a human editor who understands your industry and your business well enough to at least make your content more realistic and product-focused. But then they’ll also have to edit it for style, and once again, we’re back to where we started. Why not just pay a human writer if the AI gets the style, tone, and content wrong?
If you want to waste your time, go ahead, and try writing your content with AI. I dare you.
But over time, you’ll experiment with AI enough to come to the exact same conclusions as I’ve indicated in this article. I already did the work for you here. In fact, it took me months to realize just how inadequate AI writing is compared to human-crafted prose.
So, if you want to save time and work with an SEO agency that knows what it's doing, reach out to me at nicolas@thecerebrum.co. We'll chat about how to implement a product-driven content strategy using human writers.
