NB: This post may or may not be born out of frustration.
A while ago, I ordered a cat brush online. When it arrived, it was teeny tiny; still big enough to groom a cat, but not the size I was expecting. It was partly my fault because I didn't check the measurements and I had an idea in my head of the size it was going to be (just kind of average cat brush size). It wasn't that big of a deal, and I still used the brush, but I did feel kind of cheated. Fortunately, it wasn't as bad as ordering a piece of furniture and discovering it's actually for a doll's house.
My point is, we all know what it feels like when you don't get what you paid for or what you were expecting. And with the rise of AI writing, it's completely understandable that you may want to ensure that any writers you hire are delivering their own words. Why pay for unique content written by an expert if what you're going to get is content that you could have generated for free? It would be a waste of time and money, and it might make you feel like they've taken advantage of you. So, naturally, people are increasingly turning to tools that claim to be able to detect AI-written content with extraordinary accuracy.
These tools include the likes of Copyleaks, GPTZero or ZeroGPT, Originality.ai, and OpenAI Detector. You can find plenty of articles about whether these detectors are good at picking up AI content, some of which say they do, and some say they absolutely don't.
Do AI Content Detectors Work?
I have seen many complaints from writers of being accused by clients of using AI writing tools like ChatGPT, when everything they have written is in their own words. Some writers are even saying that accurate spelling and grammar are penalised by AI content detectors. I'm no exception to this, with AI detector tools telling me with certainty that some of my original work is at least 40% likely to be AI content (and sometimes much higher than that).
I've just put everything I've written in this post so far through Copyleaks, and it's only 53% sure it was written by a human. I'm pretty sure I'm still a human. More than 53% sure, anyway. I've also played around with Copyleaks before, changing the odd word and sentence. Changing just a couple of words means the detector goes from being mostly convinced a paragraph is entirely AI to being mostly convinced it's human, and vice versa. Removing a sentence can change the entire text from supposedly being AI to being completely human. So even if my content were AI-generated, it's likely I could easily convince the tool that it's not, just by making a couple of small tweaks.
Should You Use AI Content Detectors?
Personally, I recommend against using AI content detectors because I don't think they're accurate enough. However, if you're determined to use them, don't just take them at their word. They're not infallible, so whatever score they give you has a decent chance of being incorrect. You might use them as one of several ways to assess the quality of the content you've been given, but don't treat them like they know all.
So How Do You Catch Out Writers Using AI?
Ok, so AI detection tools don't seem to be particularly reliable. So how do you ferret out content that has been generated by AI and doesn't meet your standards? Well, I'm afraid there's no perfect method. But if you do want to find content that's right for your site, you can start by knowing what exactly it is you're looking for, and taking the time to read over the text you're given.
I know it saves a lot of time to simply plug some text into a tool and automatically get a decision on whether it's good or not. I use tools that make use of AI to check grammar or optimise content for SEO, and it can make things speedier (although sometimes it actually seems like it makes things slower). But if you really want valuable content, you need to be prepared to check it over yourself too.
The real question to ask yourself is: who is really the authority on whether content is good enough for your brand? A robot that's trying to detect the work of other robots, or you and/or your team?
Should You Care About AI-Generated Content?
Another, perhaps slightly more controversial, question to ask is: does it matter if content was written using AI? Or rather, is it as important as whether the writing meets your brief, fits your brand, and is appropriate for your audience?
I know what you're going to say. Google says it does matter if content is written by AI. But that's not strictly true. Google Search's guidance about AI-generated content actually emphasises that it's the quality of the content that matters, not how it's produced. They are looking for content that meets their "E-E-A-T" criteria, which shows expertise, experience, authoritativeness, and trustworthiness.
What Google does object to, and always has, is the use of AI to create content with the primary purpose of manipulating search rankings. Google says that AI can be used to create helpful content, but they don't want spam. The final word is that "however content is produced, those seeking success in Google Search should be looking to produce original, high-quality, people-first content".
So all of this really means that whether your content is AI-generated or written by a human (and whether an inaccurate tool says it is) isn't what's important. What matters is that your content is helpful for your audience. If it provides value, Google will like it. Of course, it's helpful to pay attention to certain ranking factors too, but SEO isn't necessarily the be-all and end-all when it comes to ranking in searching engines.
As someone who makes their living writing and never uses AI tools to write any of my content (I've yet to even use it for research, outlines, or ideas), it might seem a bit stupid of me to suggest that AI content is fine to use. But, personally, I still think the best way to end up with valuable content is to have an actual human writer create it for you. AI tools can do a decent job of following clear and simple instructions, but I don't think they're quite there yet when it comes to matching the capabilities of a person. They very well might be in the near future, but I'm still going to put my trust in people for now. And if you're thinking about using AI content detection tools, I'm guessing you are too. (But don't use them. They're tacky and I hate them.)
