Don’t put all your SEO eggs into the AI basket just yet

Geoffrey Hinton, the ‘godfather of AI’, recently resigned from Google amidst warnings of the dangers of advancing artificial intelligence technology. His comments reignited the already hotly debated topic of AI and specifically the role of Chat GPT in work and life. 

The topic opens an ethical Pandora’s Box, which we will avoid for the purposes of this article. Instead, we want to share our experience testing Chat GPT for some potential Search Engine Optimised content work. Based on our ongoing experience, we’re not outsourcing human resources to digital resources any time soon!

Let’s preface everything we’re about to say by acknowledging that from a purely content quality perspective, Chat GPT is at its worst right now. As is the nature of the AI beast, this evolutionary wheel is moving at full sprint. As such, we will likely only experience improvements, better features and more quality outcomes. 

However, don’t throw your content-writing baby out with the baby water just yet. We’re still a long way from outright replacing an experienced, knowledgeable, skilled human being. Internal linking, high-quality keyword research, optimizing content without it competing against other pages on the same website - Chat GPT is not quite there. 

However, the biggest bone we have to pick right now is the quality. Upon first inspection, it’s impressive. Dig a little deeper, though, and the warts start to show.

Like every digital media agency, we jumped onto Chat GPT as soon as possible to get a first-hand impression. One of our clients was equally quick, tasking it to write a blog post about a particular plumbing problem. 

No doubt, the client could not believe their eyes when seconds later, a full, supposedly unique blog article was submitted. They read the first paragraph, liked what they saw and sent it to us for review. 

We had to admit the opening paragraph was entirely satisfactory. But it was after the first paragraph that things fell to pieces.

We’ve all got someone in our lives who exists without a filter, right? They’re the ones we’re usually apologising quietly for in the background. Based on our experience, Chat GPT was a lot like our filterless friend. It didn’t seem to grasp what to sift out and what to leave in. 

So, possibly to tick all the boxes, it submitted an e-vomit of copy. Occasionally, something of note was written, but - oh, my days! - did we have to wade through a woodland of words to find those nuggets. 

Quantity might keep some analytical bots happy, but verbose copy that lacks quality won’t win over humans.

Did our tongue-in-cheek title make a point? 

This is linked to our previous observation, in that the flood of words included a lot of repetition. You could see how the intelligence gathered information about the same point from various sources. Rather than summarising them all, it had taken snippets from each and essentially said the same thing in five different ways. 

We knew the sweet spot for said article was around 880 words, which Chat GPT had delivered. However, once we pruned away the repetition, we were left with a 100-word introduction that couldn’t be expanded upon. 

One hundred words don’t make a blog by any stretch.

As a result of using disparate sources, it turns out the copy had inherited disparate opinions, too. So at the start of the blog, it gave advice to do ABC. However, mid-way through the copy it back-flipped on itself and said to do XYZ. Confusing and unprofessional if you blindly accept and publish.

We will be the first to admit it; plumbing isn’t the sexiest of topics, although it can get pretty dirty! 

But that’s part of the challenge for a good content writer. Finding the points of interest and amplifying them, or at the very least making a dull topic so informative you just have to stick with it. 

Based on our experiences, artificial intelligence takes a, well, robotic approach to creativity. It’s clinical; playing with nuance, introducing some colloquialism and injecting character into content seems well beyond its abilities for now.

Although this does not pertain to our experience, it is a sobering word of warning. You are ultimately responsible for anything you publish - you can’t blame AI for your failure to quality control and fact-check. 

By now, we’ve probably all heard about the infamous AI-generated copy that fabricated a sexual harassment claim against an American Law Professor and quoted a Washington Post article that didn’t exist. Not only was the allegation invented, but the major media outlet article quoted was also a figment of virtual imagination. Seriously scary stuff. Sure, this is an extreme example but it shows we can’t afford to down tools and hand all responsibility to computers just yet.

We treat Chat GPT a little like an assistant or a sounding board.

It can make worthwhile suggestions on how to lay out an article - information you need to include, for example. We also like to set a writing task after we’ve completed our work. This way, our work avoids any subconscious influence, but we get to see if AI has come up with an angle we didn’t consider. 

It’s a little like covering all bases and getting some cool and clinical peace of mind from your non-feeling, nonplussed colleague! 

How do you find Chat GPT? We’d love to hear your thoughts!