AI in Content Creation: How Should You Be Using It?

There is no question that the sudden onslaught of AI-generated content on the web is shaking up the world of content creation. The utility of tools like ChatGPT and Bard are inspiring questions both ethical and strategic:

• Should companies be using AI to generate their content? 
• Can AI do the work of our marketing firm? 
• Is our marketing firm using AI? Should they be? 
• Does AI content meet our requirements in terms of quality?

These questions have been front and center in many of our recent discussions, and we’ve spent significant time testing different methodologies for employing these tools in our work. Though we’re aware that anything we decide will certainly crumble as the rapid evolution of large language model-based tools continues, we’re willing to share what we’ve figured out so far.  

The best person to discuss this effort at Notable is Nancy Smay, our Head of Content Strategy and Managing Editor, so what follows is a Q & A with her. 

Q: With the advent of AI, are clients moving toward building more of their own content using these tools?

A: The short answer is definitely yes. But let’s look at that in a couple ways. 

First, let’s consider the point of view of our clients, who are mostly B2B companies in professional services and IT. They are definitely experimenting with various ways to employ AI to generate content. 

In some cases, they’ve built entire blog posts and LinkedIn posts in ChatGPT. And we’ve also heard that they have ended up discarding or taking down outputs when critical feedback suggested that these pieces of content didn’t fit their brand personality or tone. 

In other examples, clients have actually run our words through these tools to “tighten up” or “clarify” wording, only to lose nuance that had been carefully built into things like elevator pitches and onliness statements. 

Finally, some have told us that they’ve been using these generative helpers to spawn lists of potential topics for keynote speeches at upcoming conferences or blog topics they might like us to tackle on their behalf. 

Internally, we’ve run most of the same experiments. 

Through all this, we’ve come to a couple clear conclusions. 

What Generative AI Tools Can Do in Content Creation

Generative AI tools are great for: 

• Brainstorming lists of topics (for blogs, for posts, for speeches, etc.) 
• Tightening up language in documents where tone and personality are not critical (emails, briefs, informative papers)
• Inspiring ideas - asking questions and then receiving feedback, even if it isn’t inherently useful feedback, gives you a starting point for your own creativity. 

In many ways, these tools operate like responsive white boards. You feed in a few ideas, and they pop out a few more. The real magic lies in knowing how to prompt the tools. There are plenty of lists online to help you write great prompts.

The danger lies in relying heavily on these outputs as final products instead of fodder for exploration. That’s because these tools are not good at things like:

• Creating content that can convey or embody an established personality, whether for a brand or a person
• Utilizing nuanced language or loose analogies that can enhance relatability
• Calling on humor
• Generating unique perspectives or original content

The bottom line here is that we use these tools for quantity, not quality. If we need a lot of ideas to parse through, or we’re looking for starters for a naming project or examples of tag lines, we’ll generate a list. That said, the outputs are never the ones we end up using, but they are helpful inspiration. 

ChatGPT’s outputs are like that sourdough starter people were so obsessed with in 2020. Great things can come from them, but they’re nothing you want to consume in their raw form. 

I think we also need to keep in mind that while ChatGPT and Bard feel like your creative buddies, ready for you to pop in and brainstorm whenever, there are real people and organizations on the other side of that prompt box. Be careful what you decide to tell them, and never share personal information or a client’s proprietary intellectual property. 

Q: How do AI tools supplement the work done by content firms? Are there right and wrong ways to use this? 

A: I can only speak to the work done by one content firm - Notable.

We hit on some rights and wrongs in the previous question, but the real determinant will be the way these firms generate content in the first place.

Notable’s content as a service (Brand 360) is based on subject matter expert interviews. All of the content we build for a client comes from a super-rich, conversational discussion that we record like a mini-video podcast and then build into everything from blog posts to video clips for social. 

There have been some great developments in video editors lately, things like AI-created clips that are supposed to automatically capture the most compelling parts of an interview and create clips you can use. And while I love the idea, the technology isn’t there yet. 

We do a lot of post-production that AI can’t do yet. From creating branded frames for video to really cutting pieces that make specific points that hang on a client’s core strategic content pillars. We’ve tried hitting the “magic clip” button. It would save a ton of time. But the reality is that these tools aren’t magic. Great content still requires a human at the helm to identify that emotional resonance, the really compelling nuance that will make people want to read or listen. 

The AI-based tools that have been useful in these processes are things like Descript’s overdub and text-editing. Those things really do seem like magic to me! Auto-generation of subtitles is super helpful too, but there are always errors a human needs to correct. 

I think the wrong way to use any of these tools, whether we’re talking about text generation or video production or illustration, would be to expect to utilize the raw outputs you get back. Nothing that comes out of these is ready for prime time, and that’s not a bad thing. 

As a creative, seeing a way to implement new tools into my workflow is great, but it’s also reassuring to see that my unique skill set isn’t being replaced. 

Q. Notable works with a lot of technical companies. You said that you have experimented with using generative AI to produce content. Can you train ChatGPT to really understand tech clients? 

A: The easy answer is yes.

But the question behind your question is whether we can train these tools to create content for these clients. And unfortunately, that’s a no. 

Keep in mind that these tools do have access to tons of resources and information, so technical material isn’t out of their grasp at all. They can explain complex concepts and work out complicated equations all day long. (Though sometimes they’re wrong…)

But they can’t talk about these super complex clients in a way that will benefit the client in terms of investment or revenue, because they can’t internalize brand voice, perspective, or tone. I’ve been successful at getting a few outputs that were close, but because the tools we have now are limited in memory thanks to the reliance on tokens, it isn’t as if I can train one chat for each client and continually return to that channel having the AI retain everything we’ve discussed so far. 

So the answer, I guess, is that ChatGPT definitely can understand tech. But it cannot make the leap to talking about that tech in the specific ways a B2B tech client needs for a very specific purpose. Or at least not reliably. 

Q: Does AI change the metrics you look at to measure content success, or the prioritization of things like SEO?

A: Yes, I think it will. 

But I don’t think that the sky is falling where SEO is concerned. As with every other industry, search is integrating AI to deliver better results, which will most likely be good for those searching. The information surfaced will hopefully be more contextually relevant than ever before. 

That said, since the ease of creating generic content with these tools means there will likely be a lot more content for search engines to crawl overall, there are some lessons to keep in mind. 

Valuable, non-generic content will be more important than ever. 

If search engines are implementing AI – and in fact, the search engines own these tools – then they’re soon going to be able to easily identify content written in their own tools. It would not surprise me at all if we begin to see AI-generated content suppressed in search results. 

That will mean that search optimization will still be important, but the days of stuffing keywords have long passed, which is great news for writers. Instead of being presented with lists of words to be sure to include in blog posts and articles, writers will be challenged to create content with true substance, a unique perspective, and real value. 

It’s an exciting time to be a creative, especially a writer. In my opinion, the advent of tools to help me do my job faster, better, and more creatively is a wonderful thing.

‍