We begin this blog post by assuring you that the words you are reading were written by a real, honest-to-goodness human.
A strange disclaimer, perhaps, but an important one given the growing number of online articles written by artificially intelligent bots these days. Not just from third-rate publications, either. Sources as mainstream as the Associated Press, Washington Post, BBC, Bloomberg and many others are publishing articles written by, or at least with the help of, AI.
And while most of those articles are for fairly straightforward, templatized stories such as sports reports, financial news, and real estate trends, the mere fact that so many articles come from an AI bot instead of a journalist is cause to examine whether that’s actually a good thing. Are we really ready to trust the bots to take over the media?
The benefits to a publisher — and, as in Candor’s case, a marketing agency — for relying on computers to write articles are both clear and multitudinous: They are cost effective, they are fast, they don’t plagiarize, and they don’t take sick days.
But the inherent dangers are fairly obvious as well. Can we trust that the bots are pulling from the best sources? Do computers really know what we’ll find meaningful? Will this usher the decline of storytelling as art?
Despite our concerns, we decided to give the AI writing assistants Jasper and ChatGPT a try. Jasper and ChatGPT are two of a growing number of AI-powered writing assistants, along with Peppertype.ai, Writesonic, Anyword and more, that claim to help marketers write blog posts, sales emails, SEO content, and other marketing collateral up to 10X faster.
The way they generally work is that you plug in a bit of information about your company, its audience, the topic that you want the AI bot to write about, and maybe the tone of voice that you want it to use. It will then spit back to you an outline or in some cases a fully fleshed out blog post for you to either edit as you see fit or publish as is.
Our first impression was that the resulting blog posts were … surprisingly good. Not nearly as good as the first drafts submitted by our writers, but not so bad that they were incoherent or unreadable. Perhaps our expectations were low, but the fact that the blog posts actually made sense and weren’t completely unstructured gave us hope.
And yet, after several tries, we began to realize that the blog posts were lacking in several crucial areas. Their sentences made sense and were grammatically correct, but none rose anywhere close to the level of prose. They reported the facts just fine but were unable to connect the dots in any meaningful way. And while the tone of voice was perhaps less robotic than we had expected, it was far from fresh or exciting.
Where we found Jasper and ChatGPT useful was in outlining and organizing topics to be addressed in an article, and in generating ideas for angles and storylines to cover. It’s not that these tools provided better outlines or ideas than we could have come up with ourselves, because they didn’t, but they did provide those outlines and ideas very quickly – like, in seconds – that would have taken us 30 minutes or an hour to research, brainstorm and organize.
But where Jasper and ChatGPT fall short is in the most crucial aspects of content generation. It’s the part that makes reading content worthwhile in the first place, the part that makes us think and feel, the part that makes us happy to be alive. It’s reading something that inspires us to do our work better or to live our lives more deeply. It’s connecting the dots in a way that makes us see the world a little more clearly. It’s delivering insights that leave us somehow different from the people we were before.
Those might be lofty goals for a blog post or bylined article. Not every blog post is meant to change your life, after all. But every blog post can and should at least make you feel as if you’ve connected with another human being, if even just to smile or nod in recognition of another person’s thoughts.
And that’s where Jasper and ChatGPT fell short. Their articles felt… dry. Canned. As if they had regurgitated what they’d read but hadn’t actually learned the base material.
Perhaps the day will come when artificially intelligent bots can learn to write like humans. That’s a larger question for people much smarter than us. We lean towards the camp that feels there is something distinct about the human experience that can never be imitated.
We feel that our job — as writers and marketers — is to help our clients forge real, meaningful relationships with their prospects and customers. And for now, AI writing assistants can't do that.