Why Generative AI Can't Create Real Thought Leadership Content . . . At Least Not Yet

Last November's release of ChatGPT set off a remarkable frenzy of activity in the AI space. Over the past several weeks, a host of technology companies, from start-ups to heavyweights like Microsoft and Google, have announced or rolled out applications with generative AI capabilities.

Even now, it seems clear that generative AI applications will have a transformative impact on all aspects of business, including marketing. I've already seen dozens of articles and blog posts describing how marketers can use generative AI to improve the quality of their work and increase their productivity.

The "elephant-in-the-room" question for many marketers is whether generative AI apps can create engaging and effective marketing content and thus reduce the need for human content creators.

As you might expect, this question has triggered quite a debate in the marketing community, and there are respected voices on both sides of the issue.

Some experts argue that generative AI can produce some types of content as well as human marketers. For example, Mark Schaefer recently wrote:  "One of the common arguments in marketing is that AI can never replace the human voice. False. This technology levels the playing field and makes everybody an excellent content creator." 

Other marketing pundits contend that generative AI can't replace human content creators because it lacks human emotion and the ability to empathize. Therefore, content produced by a generative AI app can't engender an emotional connection with members of the audience. Many of these pundits also point out that generative AI can produce factually inaccurate content. 

I've been experimenting with ChatGPT and Bard (the generative AI chatbot by Google)  for the past several weeks, and my conclusion is that they can produce some types of marketing content very well, with, of course, appropriate human supervision.

For example, I uploaded a draft of this post to ChatGPT and asked it to produce ten social media posts (for LinkedIn and Twitter) based on my draft. The posts created by ChatGPT were, on the whole, just as good as I would have written.

There is, however, one type of marketing content that the current incarnation of generative AI applications can't produce.

Thought Leadership Content Is Different

Generative AI apps can't create thought leadership content because true thought leadership content must meet three core requirements.

• Relevance - True thought leadership content must provide insights that are relevant to its target audience, and the best content will address topics that can have a major impact on the business success of the target audience.

• Authority - Real thought leadership content is authoritative. The information provided by the content must be supported by sound and persuasive evidence.

• Novelty - Merriam-Webster defines the word novel as "new and not resembling something formerly known or used." True thought leadership content provides insights that add something new to the body of knowledge about a topic that the audience can't find elsewhere.

These three requirements are equally essential because real thought leadership is like a three-legged stool. And we all know what happens if you remove or break one leg of a three-legged stool.

Real thought leadership content is always based on information and insights that did not previously exist, and this explains why generative AI applications can't create true thought leadership content.

Generative AI apps such as ChatGPT are built on large language models, and these models are "trained" using vast amounts of data (i.e. information) published on the internet, in books, and in other sources.

When you ask a generative AI app to create a piece of content (e.g. a blog post), it bases the content on your instructions (a "prompt") and on the information it was trained on. A generative AI app cannot draw on any information that wasn't included in your prompt or in its training data. This makes it impossible for a generative AI app to create real thought leadership content.

In one of his great Almost Timely News newsletters, Christopher Penn succinctly captured this limitation when he wrote:  "More often than not, they [large language models and generative AI] create mediocre to good results, but rarely excellent and never unique, never-before-seen results. That's because their nature is an averaging of what already exists, tuned to deliver above average results."

None of this means that generative AI has no role to play in the development of thought leadership content. For example, I've been using ChatGPT on an experimental basis to vet possible thought leadership topics.

To develop novel thought leadership content, you obviously need to avoid topics that have already been adequately discussed by others. I've been using "conversations" with ChatGPT to get a preliminary indication of whether a possible topic has already been widely addressed. In essence, I'm looking for topics that ChatGPT doesn't know much about - or doesn't know specific things about.

The bottom line is, content must meet a high standard to be considered real thought leadership content. The generative AI applications available today have amazing capabilities, and the apps that will be available in the near future will likely make today's apps look rather simplistic. But, the development of real thought leadership content will require human creators for the foreseeable future.

Illustration generated by DALL-E 2.