AI content tools are getting a lot of attention in the marketing industry right now – some of it good, some not so much. There’s also a lot of confusion surrounding AI tools, in general, mostly stirred up by promotional press coverage.
In this article, we discuss whether AI-generated content can rank well in search engines by assessing how AI content tools work, the quality of the content they produce and their ability to help you produce content faster.
In this article, we answer the following questions about AI-generated content:
AI-generated content is created by algorithms using a variety of machine learning technologies. Currently, most AI content tools use a machine learning model called GPT-3, developed by OpenAI, which uses artificial neural networks to perform deep learning.
Essentially, this means that, after extensive training, the algorithm is capable of learning and improving by itself.
Why is this important? Well, it’s worth understanding how these algorithms work if you’re interested in using AI content tools. The idea of algorithms learning by themselves sounds incredible but it’s easy to overestimate their ability.
Machine learning models like GPT-3 require vast amounts of human-assisted training to reach the point of self learning. For complex tasks like language processing, this involves huge volumes of data and millions of repetitions until the algorithm succeeds and fails often enough to achieve acceptable results without human intervention.
Once a machine learning model achieves the benchmarks set out by its developers, it’s ready for implementation or release. In the case of GPT-3, OpenAI released the model as a commercial API, allowing third-party companies to implement it into their own products – for a healthy fee, of course.
Now, if you try out an AI content tool, it’s most likely using the GPT-3 model.
Most of the popular AI content tools getting attention now use the GPT-3 machine learning model, including all of the following:
Each product has its own interface and set of features but the core technology underneath is essentially the same.
This means the quality of the output you’ll get will be very similar, too. In other words, if you’re trying out several different AI content tools, don’t expect any of them to produce significantly better content. So, let’s see what happens when we create a blog post using one of these AI content tools.
To generate content with any of these tools, you have to provide some keywords and basic information for them to work with. The tool we’ve selected has a dedicated feature for blog writing (most of them do) and it requires the following information to get started:
We want to test the tool as much as possible so we’ll simply define the topic as “blog writing” and the keywords “blog writing,” “content marketing,” and “SEO” while leaving tone as the default “Friendly”.
Once we’ve provided this information, we can move on to the next stage where the tool asks us to create an outline for the article.
The tool wants us to create an outline for the blog post, which simply uses keywords to define different sections for the piece. The tool has automatically generated an outline for us with the following four sections:
Clearly, those aren’t acceptable topics or headings for a useful article on blog writing. We can regenerate these suggestions by clicking the Regenerate All button and its second effort comes up with a generic list of benefits:
This is better than its first attempt but it’s still too generic and lacking any cohesive structure. We could keep hitting the Regenerate All button and hope something eventually works out (it doesn’t, we tried) but this isn’t a practical workaround.
We need to intervene here and give the tool a reasonable structure to work with. On the spot, let’s give the tool these sections/headings and see what it can do:
Once we’re done, the tool says it’s generating our talking points for the blog post so let’s see what it comes up with.
For each heading, the tool proposes several talking points, which it generates from relevant content around the web. Here’s what it came up with for the first section:
Again, you can regenerate all of these, add more talking points automatically or add your own custom points for the tool to work with. To give you an idea of what else the tool comes up with, these are its suggestions for optimising blog writing for SEO:
As you can see, it has successfully matched keywords with the topic and produced a list of accurate tips. However, it’s failed to touch on any of the other aspects of optimising blog posts for SEO.
Next, the tool generates content for each section, based on the topics and talking points confirmed in the previous steps.
The tool successfully ties blog writing in with content, readers, keywords, link building and several other relevant sub-topics. It also picks up on the X vs Y nature of the heading and provides comparative definitions for blog writing and SEO.
The content itself doesn’t make much sense but it has managed to match topics and phrases reasonably well in this case.
If you sign up for an AI content tool and expect it to produce quality at the push of a button, you’re going to be disappointed. Unfortunately, a lot of the marketing around AI content tools is misleading. In reality, these tools aren’t designed to create original content from scratch; they’re designed to help writers produce content faster.
AI content tools can help in several ways:
Rather than starting with a blank page, AI content tools ask you to provide a title, add your keywords and structure the post before you get started. If you’re not used to doing this with every piece of content, these tools will help you get into the habit.
They’ll also suggest some relevant topics to cover but, for the most part, these will be very generic recommendations – not the kind of stuff you want to include in detailed or technical posts.
In terms of the actual content these tools generate, quality is fairly low. At best, it produces fluff content with occasional hints of relevance but never achieves enough clarity to say anything meaningful.
Let’s go back to our AI-generated content for a moment.
First, the tool provides a definition of blog writing that makes sense but doesn’t really say anything. “Blog writing is about creating content for readers” is like saying “photography is about taking photos”. Technically, it makes sense and it describes one specific aspect of the subject but it doesn’t offer any valuable information.
The tool struggles to expand upon its own point in the next sentence. It pulls in phrases like “organic” and relevant terms in “keyword research” and “link building” but fails to achieve any clarity.
Two sentences into this piece of content and it’s offered nothing of value. However, you can start to see a point vaguely develop in the second paragraph.
Compare the first sentences from each paragraph and we’re told that blog writing creates content for readers while SEO optimises content for search engines. The accuracy of this statement is debatable and the AI-generated content takes too long to make it but these two sentences could help you get started with a section comparing blog writing and SEO.
It’s not going to do all of the work for you but it could help you get started.
The second sentence in this paragraph does a reasonable job of explaining the role of keywords and links in SEO. The issue is it reduces the broad field of SEO to keywords and links when there’s so much more to talk about, but you can easily edit this sentence and add to it.
The final paragraph in this block of AI-generated content is the most problematic.
Again, it pulls in plenty of relevant phrases but simply mashes them into one long sentence that doesn’t really make sense.
Worse still, the point it almost manages to make contradicts the purpose of the article itself by downplaying the importance of the subject we’re writing about. It also trips itself up grammatically with “when it comes time to” and fails to match the British spelling of “optimise” that’s included in the heading for this section.
This paragraph offers nothing useful so you would have to delete it and write something from scratch.
It’s easy to criticise the output of AI content tools but expecting them to match the quality of professional content writers is unfair. Keep in mind that these algorithms can only generate content by analysing existing content from around the web. All they’re really doing is mashing phrases and sentences together from patterns they spot in relevant content covering the same topics.
There are several issues with this:
Even if these machine learning models could match human writers linguistically, they’re not going to produce anything new or insightful. The technology will continue to improve, however, as things stand, even the best AI content tools struggle to produce coherent sentences or make any meaningful points.
It certainly can’t explain anything technical or go into detail about a subject – not without extensive human intervention.
Earlier this year, Google responded to the increased use of AI content tools with a clear statement. More specifically, Google’s John Mueller responded to a question in an April 2022 SEO Office Hours session asking about Google’s view on AI content tools.
Here’s the short version of his answer:
“Currently it’s all against the webmaster guidelines. So from our point of view, if we were to run across something like that, if the webspam team were to see it, they would see it as spam.”
Before reaching this conclusion, Mueller provided a more detailed answer:
“For us these would, essentially, still fall into the category of automatically generated content which is something we’ve had in the Webmaster Guidelines since almost the beginning.
And people have been automatically generating content in lots of different ways. And for us, if you’re using machine learning tools to generate your content, it’s essentially the same as if you’re just shuffling words around, or looking up synonyms, or doing the translation tricks that people used to do. Those kind of things.
My suspicion is maybe the quality of content is a little bit better than the really old school tools, but for us it’s still automatically generated content, and that means for us it’s still against the Webmaster Guidelines. So we would consider that to be spam.”
Mueller acknowledges that the latest breed of AI content tools are capable of producing better quality than some automation technologies in the past. However, he’s clear that the output of AI content tools is still considered web spam and it’s hard to agree when you consider the quality we looked at in the previous section.
The standard output you get from AI content tools isn’t going to rank well in search engines. More importantly, it’s not going to offer any value to your target audience or convince them to take action.
Technically, the content these tools produce can get indexed and rank in search engines. It’s not going to pass Google’s quality guidelines or build any traction, though. Users won’t engage with this content, publishers won’t link to it and Google could penalise you if it discovers you’re publishing AI-generated content.
However, this only applies if you’re using the content these tools generate without editing and improving it.
Nobody is saying you can’t use these tools to speed up content production, though. As long as you’re editing the output of AI tools and ensuring the content delivers quality and value to your target audience, you’ll have no issue ranking well in search engines.
The question is how much time these tools can really save you. You have to spend so much time editing and, in many cases, rewriting their output that the productivity gains can be minimal. More than anything, these tools may come up with vague ideas to help you get started on a specific topic or section of an article.
It’s difficult to answer this question for everyone. If AI content tools can help you produce quality content faster and the productivity gains are worth the software fees, then go ahead. Just make sure you edit the output enough to make it original and valuable to your target audience.
As with many AI tools, the reality doesn’t always live up to the hype and it’s important to understand the limitations.
As AI content tools continue to improve, the balance may swing further in their favour. They don’t need to match the quality of professional content writers; they simply need to speed up content production enough to help teams achieve more with their existing resources – the same as any automation technology.
If you need help producing better quality content or producing more of it, the current breed of AI content tools probably isn’t ready to step up. Luckily, our content team is always ready to help and you can call us on 023 9283 0281 or fill out this contact form to get things moving.
Josh is an SEO specialist who joined Vertical Leap at the beginning of 2022. He studied journalism and was a reporter for a local newspaper before moving into digital marketing to combine his passion for the written word with his love for all things data-led. He believes content is king and that the key to success in this highly competitive landscape is to be uniquely useful. His main passion outside of work is skiing and he spent six winter seasons in the French Alps.
Categories: Data & Analytics
Categories: Content Marketing, SEO