Is Content Marketing still a profession worth pursuing, in the age of AI? Let’s deep-dive into the effects of AI on how we interact with the internet.
Let’s start by looking at the good ol’ search session.
Why do we search?
We search because we want to know something.
In the 90s, talking about the internet as a physical place was science fiction, but today, the reality is that the majority of entries into Google are what we call “navigational” searches.
People just want to get from place A to place B. This is one of the major uses of Google, and many content marketers get tripped up thinking that the navigational keywords will bring them some customers.
Here are some giveaways that your beautiful keyword is just navigational:
- The keywords are usually low KD and don’t have any content ranking in the top 10.
- The volume is in the bazillions
- Related keywords have something like {keyword} + login, or {keyword} + portal, app, etc.
Navigational searches aren’t that hard to identify. They are the type of search you want to avoid targeting in content marketing. They’re important because they show a critical flaw with LLM-based AI as we currently know it in 2024.
When you use navigational queries with LLMs, they provide a link that sounds right but needs to be corrected. That’s bad news for the large portion of searchers using Google to get from point A to B. But…
Navigational searches are probably going to stick around. Here’s why:
ChatGPT doesn’t know where things are, just where they might be.
ChatGPT doesn’t have good relational reasoning. At all. Have you ever tried to get it to do math? It might drop a variable partway through, or say 2+2 = 5.
ChatGPT is supposed to be good at coding, which relies heavily on a lot of short-term memory and strict relationships. But have you noticed that if you shift the focus to a new problem, it will morph code to drop entire sections?
Maybe you haven’t coded with AI, but you still may have noticed this. You start talking about one subject, then switch to something slightly different, and it begins critiquing you on ideas it came up with earlier.
This has to do with the context window. The context window is the number of tokens (which you can think of as words) that an LLM can juggle at once. Anything outside of the context window won’t get factored into responses, and within the context window will be emphasized.
Compared to humans, even the most state-of-the-art LLMs have a tiny context window (only about 2 million words). Human memory has been estimated at roughly 2.5 TB, which runs about 1 million times the size of the largest LLM context window.
So LLMs can’t sort through a large amount of information that isn’t part of their initial training, whereas humans have relatively unlimited context windows. If you’re trying to write about a new topic or domain that doesn’t have a lot of text written about it on the web, guess what? AI doesn’t have a clue about it.
Meta’s AI chief says the best LLMs still aren’t “smarter than a house cat” – The Observer (2024 02) This is exactly because of this lack of direct, relational reasoning. LLMs use associative mapping for all of their outputs, which means they are essentially dreaming up everything they say.
In essence, they don’t actually know anything, they can just guess really well.
What ChatGPT does badly is an enormous opportunity for Content Marketers.
ChatGPT and other Large Language Models have flaws, and that’s where humans writing helpful content can do better than AI.
Again, we search because we want to know something.
Google AI can tell us the answer to most simple questions, and perhaps we walk away happy, and that’s the end of our searching experience. But what happens when we want to know the answer to something really difficult to answer?
Helping an inquisitive person navigate their own rabbit hole is something Google’s AI can’t do. It provides quick, simple answers to common questions. What happens when the question is answered best- by another question?
From Content for Content’s Sake to Strategic Marketing
AI gives us incredible abilities to answer easy questions, even easier than before, but for content marketers, is it worth the effort to make such pieces of content? When we create this content and rank No. 1, Google creates a “featured snippet” so users don’t even need to click through to see the answer.
So-called ToFu (top-of-funnel) content might look like this:
- “How to write a blog post with ChatGPT”
- “5 Amazing Capabilities of ChatGPT Most People Don’t Know”
- “ChatGPT and Creative Content: The Ultimate Guide in 2024”
People read the heck out of these types of content, and blogs have the ability to crank them out faster than ever before. ChatGPT can make this kind of content fast, cheap, and high quality.
But the reason that AI can create this content so incredibly well is the same reason why perhaps you shouldn’t make it.
ChatGPT and other LLMs excel at creating content that is similar to its training material. That means there’s a lot of this content already out there.
The type of content that doesn’t exist is the content that takes users’ situations into account. It answers the question “for me, for my use case.” This is the kind of answer that AI utterly fails to produce.
Shifting Focus: Helping Users Make Final Comparisons and Decisions
Deep in the rabbit hole is the place you should start as a content marketer. In fact, the place you can have the most impact in the buyer journey is at the end. Buyers make their final decisions after they’ve weighed their “why” against the product’s “why.” If the customer decides there’s a fit, they buy.
Buying doesn’t always mean literally buying. It’s just how we refer to the point where a buyer decides to go forward with your product to the next stage.
It can be setting up an account for a free trial, clicking through the onboarding process, or registering for a call with sales. As a content marketer, you should be that specific when making comparison content.
So the end decision isn’t whether product A or product B is better, or what are the feature differences. If you have feature landing pages for your product, AI can likely help the user figure this out.
Instead, you should focus on the first moments of engagement for the product. What will it be like? How is it different than your competitors’ products?
And you don’t have to stop making content at your competitors, either.
Buyers are likely wondering-
- “Which plan is right for me? Will I be ok with Starter, or do I need some feature in Business?”
- “Why is the webinar an hour long? Will I see how it works enough to decide if we can use it for our use case?”
- “How can I roll this out to my team? Will the onboarding be enough, or do I need to develop training or host a workshop for the team? Do I need to buy the software to know how to set it up for our needs, or can I figure it out from the support documentation?”
These are all questions that AI can’t come up with easily. Neither can you as a member of the marketing team! To truly answer these questions for customers, you’ll need to do a lot more:
- Put on more hats. Content Marketers should think like sales, devs, products, and customer success, not just marketers.
- Talk to more people. You need to talk to the founders, devs, and sales. Then, talk to CS, Product, and Community. These people have the missing parts of the puzzle that you need.
- Bring the Voice of Customer. Get user testing data, build relationships and case studies, and interview (or be a part of interviews) with customers often.
I would argue that generative AI can’t do any of the things above. Great content marketers have known this all along, and may not change much about their day-to-day. Good content marketers will need to elevate their thinking and realize that content marketers must break out of the “marketer” silo to be great at their jobs.
Strengths of AI in Content Marketing
LLMs and generative AI are a force multiplier. If your content consists of listicles, how-tos, and ultimate guides, you can crank these pieces out 10x faster than before.
And if you write cross-functional content based on interviews, real scenarios, data, and personal experience, you can write these pieces 10x better than before.
Content marketers who avoid AI are wasting time. It is a must-have tool.
Just not for writing content.
If ChatGPT is turning out great content for your subject, your subject is wrong.
LLMs can only write cohesive, complete content on topics that have been extensively covered in their training material already. That means someone else has already written (a lot) about the topic matter.
Your job is to go deeper, to go where the waters are murky, and to be a guide for people who find themselves asking the same questions.
What LLMs lack in rational capability, they more than makeup for in associative capability.
They can also write code, enabling drill-down analysis by humans on large datasets.
Sounds great, but I’m not a data analyst, I’m a content marketer! Right?
Wrong. We are all data analysts now too.
We’re also AI engineers!
Try this prompt in ChatGPT:
Please see the exported csv of all comments on social media for the last year, and other csv of our blog schedule from Asana. Now, take a look at our competitors’ a, b, and c blog posting schedules.
Please perform initial EDA on these datasets, join them into a single table, fill in missing elements, perform qualitative to quantitative conversion, and train a number of machine learning models to predict seasonal changes in topical popularity, using best practices for exploratory data analysis.
Oh, and provide me the code to run these models myself and visualize the results using Plotly or a similar open source visualization library in python. Provide individual code blocks that I can easily copy and paste directly into Google Colab.
As a human, you are the driver of an enormously capable machine, able to accomplish analysis tasks like this with ease.
In the same conversation, now you can ask it to identify common questions and topics that your competitors cover on their social media, but your product doesn’t. All in the span of 15 minutes.
AI can write like you, but even more personal.
Becoming a great interviewer is probably the number one soft skill that any content marketer can have. Because with ChatGPT, you have another person you can interview.
That’s right, you can interview yourself!
Asking ourselves questions is something I have found that the human brain is notoriously bad at. To quote HBO’s “Westworld,” we’re all stuck in our own loop. AI can change that- and all it takes is a simple prompt:
I want you to take the role of a prize-winning journalist, who built a distinguished career as an expert in {X}, before taking on a mission to ask the most critical, hard-hitting, insightful questions and expose the real truth of {X}. Now, start interviewing me free-form, and don’t be afraid to challenge me and get the best story you can, by getting me to really open up.
(replace {X} with your topic)
I highly recommend turning on voice mode with ChatGPT in the mobile app for this part, especially the new GPT4o model.
As a ChatGPT team member, which has approximately twice the response cap of plus users, GPT4o can now have 160 responses every 3 hours (as of May 2024). This means you’ll never run into a response limit. In my experience, you’ll get hoarse and tired of talking first.
And your interview could go on for a long, long time. People don’t utilize this fact enough.
As ChatGPT gets to know you better, the interview will get more personalized. Don’t get discouraged if it asks a couple of irrelevant questions. Just say it’s not relevant to me, and move on.
The context window for ChatGPT 4o is 128 thousand tokens. This is enough for a typical interview to last for over 80 days!
What’s the point of these super-long interviews? They are to get all intricacies and subtleties of your ideas down into one place, before writing your content that harnesses your voice and ideas.
After talking to ChatGPT for an hour or so, ask it to create an outline of a piece of content deep-diving into one of the topics you’ve discussed. By this point, you’ve provided enough subject matter within the hour to make a piece of content truly unique.
And not only will your content be unique, but it will sound like you. After an hour of interviewing, ChatGPT4o will have captured a large number of idiosyncracies that make your voice you, so you can have it written in the way you want to sound.
I call it more personal than personalized because it can replicate your subconscious style of expression, organizing your thoughts in a distilled, concentrated way.
So, to wrap up
Don’t discount AI in your everyday work. Let it elevate what you do as a content marketer. Focus on the Why, and make it your mission to discover the why for your business if other departments haven’t found it yet.
You are an analyst now. Upload your data, and get it crunching. All you have to do is ask the right questions, and the insights are there.
Use the space. GPT4o has more context and more responses per hour than GPT4. If you use this additional room to spread out and spend more time on a subject before moving to the writing tasks, you’ll have a much better result.
Leave a Reply