Details
The explosion of Gen AI following the release of Chat GPT in late 2022 has rightly been seen as a seismic moment for marketing. The years since have seen an explosion of use case studies and applications that seem to herald an entirely new era of productivity and automation.
But as the dust settles and content marketers take stock, it has become clear that AI is not the panacea that some have claimed.
No one can deny that there is huge potential. By analysing huge amounts of data almost instantaneously, AI tools can help to create highly personalised, customised experiences that are calibrated to the needs of audiences. They can also optimise content activation and placement, helping to ensure that content reaches its audience at the right time and in the right place.
More and more companies are using AI tools to create content. The speed with which this has become the norm is extraordinary.
Although it’s hard to come by data on just how much online content has been created using AI tools, some estimates suggest around 10%. Often, content may not be created using AI but translated with AI algorithms. One study from Amazon Web Services estimates that 57% of internet content has been translated using an AI algorithm, and this is usually poor quality.
Obvious appeal, huge downsides
Compared with the traditional approach of using good old-fashioned writers and creatives to produce content, relying on AI is of course incredibly cost-effective. Content marketers and native advertising professionals can create content at scale in a way that is almost effortless.
The appeal of these tools to agencies and media organisations is huge when you consider that 45% of marketers say they spend about half of their week creating marketing content, according to the AI Marketing survey from Capterra.
Yet despite the obvious appeal, there are huge downsides.
The quality of AI generated content is only as good as the data on which the model has been trained. It is by definition not original, and is ultimately derivative of other content sources. In a world where discerning audiences expect content and thought leadership to tell them something new in order to be deserving of their attention, this means that most AI content will fail the basic test of being able to engage the audience, let alone influence them or change their thinking.
Quality issues with AI content are well known. In addition to the risk of bias, hallucinations and breach of intellectual property, some researchers have pointed to the potential for so-called “model collapse”, a degenerative process whereby AI systems trained on their own outputs get stuck in a doom loop of declining quality, with errors and inaccuracies becoming embedded and magnified over time.
Regression to the mean
Content from large language models also tends towards a “regression to the mean”. If you know little about a topic, then it can be incredibly useful in raising the sum of your knowledge to a medium standard. But it is less helpful for true experts or in highlighting information that is at the edges of current knowledge and that could expand the frontiers of what we know.
Again, discerning readers will expect more than an average standard - and this certainly fails the test of whether something could be considered truly groundbreaking or genuine “thought leadership”. Attaining that level of quality currently has to come from human input.
There are other issues. Good content is more than just information – it must also incorporate an essential emotional element that helps to engage audiences with empathy. This is currently lacking from AI generated content, which tends to feel somewhat soulless because it relies on logic and algorithms.
These shortcomings, for now, maintain a “competitive moat” around those companies and media organisations that invest in and create the very best content.
By combining original insight, as opposed to derivative thinking, and adding a layer of human, emotional intelligence, the best writers and producers have an important advantage over AI, at least for now. They can connect with audiences in ways that AI tools cannot and what they lack in logic and brute force computing power, they make up in terms of being able to build a strong and lasting emotional connection.
A race to the bottom
The gap between AI and human content is real - although not everyone is able to spot it. The AI Marketing Survey from Capterra found that 82% of content producers thought that the quality of AI content was as good as that created by humans.
Now, this surely depends on the type of content being discussed. Maybe AI tools are fine at producing basic email copy or product information. But high quality content such as thought leadership or native advertising just cannot be created effectively using AI as things stand.
The gulf in quality between AI and human generated content does not seem to be a deterrent for marketers. Many have become addicted to the time-saving properties of AI, without necessarily considering the downsides. Yes, these tools make content production easier and can ‘write” content on the back of a simple prompt in a matter of seconds. But the quality is simply not there.
Moreover, there is nothing proprietary about this technology. The same tools are available to everyone, which means there is very little sustainable competitive advantage from using them. There may be a fleeting, short-lived one for the first movers, but this quickly evaporates as competitors discover and adopt the same tools.
Without any real competitive advantage, we’re likely to see a race to the bottom, with downward pressure on agency pricing as clients realise that they are no longer paying for human expertise.
AI tools can improve productivity and content producers are right to be experimenting with them to speed up repetitive tasks, help with research or spark new ideas. But an overreliance on them for content production is highly risky. These tools will improve over time and their value in producing new content will increase.
But for now there is nothing that can beat the skills, creativity and empathy of human writers, editors and designers.