Replace me? Not so fast.
Let’s be frank, your content writer is dangerously close to being more hassle than they’re worth: Hard to get ahold of, expensive, a bit of a prima donna when you attempt to change their copy and, last but not least ALWAYS pushing their deadline (this blog entry is no exception, 2 days behind and counting).
And all that management hassle comes before we even get a chance to determine if the work they produce is actually found useful by your clients and getting you search ranking.
As an organization that specializes in content quality, we’re always looking for the best way to communicate more clearly and effectively with customers – be it through a more effective Q/A process, streamlining workflows or taking advantage of new technology that promises to eliminate some of the unpredictability of human-powered content creation.
The rise of the machines – and the fix to all those human problems.
Artificial Intelligence (AI) Copywriting software like Jasper and AI content creation tools like Chat GPT offer the promise of speed, ease and economy in generating the website copywriting you need to stay relevant in the industry.
But, before we give that failed-novelist-turned-copywriter their walking papers we need to consider some of the limitations of AI Because, just like the industrial revolution that did so much good for quality of life, it also comes with some complications and concerns.
This human writer is impressed with the potential available when it comes to AI Copywriting and the integration of man and machine could have a powerful effect on your communications strategy.
AI Copywriting 101: What it is and How You Can Harness It
If you’ve followed the development of AI models and just want to understand the current limits to AI content creation, you can skip the next few paragraphs.
But if you’re not familiar with these programs, you’ll better understand the limitations once you understand how AI fits into the world of productive content creation.
AI As the Assistant To The Creator
Man has always been, by nature, a tool developer – constantly experimenting to create newer, easier ways to complete a task. A tool does not replace the user, it merely makes their work easier and even perhaps changes the nature of their work all together. Driving an ox to plow a field is a very different skill than driving a tractor but the change in the technology allows for more production and better quality of life.
In the field of translation, we have been seeing this reality for the past 20 years with the improvements to machine translation and computer assisted translation tools. The job of translator has not gone away despite the fears. Translators have instead been able to take advantage of these new tools and technologies to improve their speed and quality and increase customer satisfaction.
Cyborg 1.0: The Search Engine
To frame the evolution of AI writing, let’s step back to the last most related tech advancement: The search engine.
Philosophically a search engine is an intelligence multiplier. It gives the broad-minded generalist a tool to “zoom in” on nearly any topic and become a “temporary expert” long enough to accomplish a task then zoom back out and move on to the next opportunity or challenge.
As an example: Armed with the basic knowledge of auto mechanics (the generalist base), a few keystrokes will render an entire step by step tutorial on replacing the timing belt on your exact car (year, make, engine sub model, you name it) – all the specific knowledge is “learned” by the search engine.
With a few more keystrokes you can plan and plant a colonial “kitchen garden”, learn the tariff rates for importing cotton t-shirts into Canada from India or learn about Somatic experiencing as a psychological treatment for the trauma left by a car accident.
Each of those individual topics would have taken weeks or even months of research (library searches, book skimming, copious note taking) 20 years ago – and yet now that expert-level knowledge is available in seconds thanks to ready access to the internet via our electronics.
As Elon Musk boldly proclaimed recently, our attachment to our smartphones and the intelligence augmentation they offer has already effectively made us into cyborgs. And the only limit to your knowledge potential now is your will to search the topic and your brain’s ability to sort, process and integrate the results on offer.
Cyborg 2.0: The Leap From Machine Search to Machine Integration
From a creation standpoint an AI copywriting tool is the next step up on the content ladder.
It promises the ability to compile those search results, intuit context and stitch them together into a narrative on any topic using the same machine logic we’ve been using to create computer programs for years.
If you identify the output format of the narrative – say “website copy” or “blog entry” – the AI can serve as a ghost writer to generate the content in the exact format you need.
It’s a truly powerful tool, but, just like the search engine, AI copywriting tools have limitations.
5 Limitations and Concerns to AI Content Creation – Or Why “Skynet” Won’t be Taking Over Quite Yet
1. Search Engine Rank and De-rank: Will the Machine Turn Off the Other Machine?
As we all know, web content has two audiences: The human’s looking for relevant knowledge and the search engines rating and ranking.
Your AI generated content may in fact answer all the related questions a prospect is searching and may hit all the keywords correctly. But the ease of reading and crafted voice will still be considerations for the ultimate target – the human reader.
There are already academic tools that can detect the percentage of content that has been written by AI While this may not seem like an issue in the commercial space, the long-term concern is that search engines will use the same algorithm to de-rate machine-written articles as the marketplace fills with content.
We simply don’t know if or when this could become a concern and so, having a communications strategy too heavily based on AI content could lead to a huge drop in traffic if those adjustments are made.
2. Big Fish, Small Pond? The Limitations to Machine Learning
Because AI writing tools are essentially crowd sourcing your content, they are limited to the knowledge that is already posted on the internet. That means expressing thoughts that take a truly novel approach to a problem is a serious challenge.
Of Timespans, Data Parameters and the Limitations They Hold
Jasper AI claims to have scanned a full 10% of the massive amount of information in the internet and Chat GPT currently uses 175 billion parameters in its learning data set and a historical knowledge that extends to September 2021.
While that’s impressive for the amount of data these models holds, it poses an obvious challenge for generating content on fast evolving issues.
For example, a blog post on the War in Ukraine and its forecasted effect on commodities prices is not a topic Chat GPT can analyze. However, one could still use the tool to help formulate historical context to set up the post.
To Err is Human and, Therefore, Also AI
Just like a Wikipedia article you read can have an error and google search can yield bad information, AI can use this incorrect information to reach a wrong conclusion.
Recent testing pitted Chat GPT against the American Bar Exam and, while the professors who conducted the experiment noted the AI produced sophisticated responses to the questions, it didn’t pass the overall exam: The AI got just over 50% on the exam questions right compared to the average human score of 68%.
A Bad Learning Habit on Repeat: When Errors Create Errors
Bad datasets feeding bad conclusions also creates a potential for a repeating error loop as the incorrect information feeds into the next round of content creation. In short, what happens when AI learns incorrect information and then builds on it to create more bad information?
To remedy this looping conundrum, each AI generated article will need to be fact checked extensively to make sure you are publishing correct information your prospects and customers will find valuable. Yet vetting AI generated answers can be a challenge because the outputted product does not include source notes or citations.
3. Return of the Clone Wars – Does Brand Differentiation Suffer with AI?
What happens when all your competitors are all using the same technology? Stalemate.
And while AI is a breakthrough full of potential for the early adopter, as its use becomes more widespread the danger of providing the same information, offering the same solutions and having the same tone (all lovingly crowd-sourced from the internet via your AI software) becomes very real.
As any marketer will tell you, differentiation of your product is a must and communication style is one of the fastest ways to set yourself apart while searching for prospects.
In a competitive, fast-evolving marketplace this differentiation is a key to survival and we simply don’t know how well AI tools will be able to do that yet.
How communications teams harness the immense opportunity of AI tools without committing “brand genericide” may be one of the most exciting areas of study for marketers over the next decade.
4. Professional Quality Content: The Human Details Make the Difference
Author and educator Daniel Herman wrote an article for the Atlantic Monthly Magazine recently where he “graded” content from Chat GPT and found it to be better in quality than most High School students’ papers.
That’s a laudable result for machine learning, but, one has to ask, as a business professional, would you trust your company’s email copywriting or blog content to the communications skills of most high schoolers?
That’s not an insult, but rather a caveat to relying too strongly on AI.
In another cautionary tale, consumers surveyed by a company who offered customer service chatbots were less than pleased to find out that 72% of survey respondents found chatbots to be a waste of time.
Based on these results, the question posed in Mr. Herman’s previously cited article bears mentioning:
“Is this moment more like the invention of the calculator, saving me from the tedium of long division, or more like the invention of the player piano, robbing us of what can be communicated only through human emotion?”
5. AI Public Perceptions Bias: The Tool Nobody Wants to Admit to Using
Recently online tech news outlet CNET was “caught” using an AI technology to generate a series of 70 “financial explainer” articles.
The response from the public was swift and the staff at CNET was forced into damage control mode trying to explain that it was a tech test and promising to perform a full audit of the material to guarantee no errors were found that could lead readers to make poor financial decisions.
No doubt AI generated content has a stigma to it and it’s quite amusing to see all the editorial ink online covering the topic of Chat GP – but all going to great length to insist the stories about Chat GPT were not written actually using Chat GPT or other AI software.
Which points out an uncomfortable truth: Everyone is excited to take advantage of the potential time and cost-saving opportunity of AI – but nobody wants it used on themselves.
So, if you want to take advantage of the benefits of AI to generate messaging, it’s important to be transparent about it with users. Explaining that AI speeds up the communications process, allows you to deliver more customized communications and keeps them up to date on communications is far preferable to trying to “sneak it past” them. In the era of social media honesty and openness is always a better communication strategy.
Limitations Now, But the Race for Refinement is On
In a sense the testing process for AI copywriting will be the same as it has been for all communication: Can it stand the culling that measures work on its merit to the reader and rewards the best communicators?
Great copywriting has something special to it that communicates that much more clearly, invites understanding that much more readily and is just plain fun to read – even when covering a seemingly mundane topic. The great admen rise above the noise using their wits and the tools at their disposal.
And, just like a calculator mentioned in the quote above, the prospect of using AI as a tool to speed up the communication generating process – without compromising the human elements we all crave is where the future lies.
Stay tuned, because the first chapter of the AI saga has only just been written.
And, of course, if you would like to assess and improve your content in any language, drop us a line. We’d love to chat – human to human – about what the machines can do to help us all improve our communications game.