Let's kick things off with the current state of artificial intelligence in content marketing. The questions in this section focus on the extent to which AI tools are already being used by brands and agencies.
Currently, more than 63% of content creators suggest that they are not using AI tools and in my opinion, this presents the greatest opportunity for other content creators who are embracing AI tools. The reason is simple, with AI you have the ability to be more efficient with your content creation efforts.
Considering the very recent rise of advanced AI content models, these results are encouraging for the AI content industry. I expect these numbers to go up as the development of more advanced AI content models progress and new technologies are released to the public.
Really interesting this. When I surveyed agency owners I got a similar result. 29% had used in production, with 65% having used these tools in some way (including personal experiments, test projects etc).
I would expect Aira's audience to be skewed towards marketing and SEO than my more general agency audience, and I would fully expect usage of generative AI tools to be higher in those segments.
I don’t, because there isn’t (yet, that I know of) a copywriting AI that can create actually-funny jokes. There is one built by the former head writer for Letterman, Joe Toplyn, though its results are hit-or-miss.
More importantly, the last time I experimented with using AI for a draft, the thing MADE UP names and facts out of thin air. That’s… not ideal if, say, one of your goals is to avoid contributing to the overwhelming flood of mis- and disinformation being published daily.
I wonder to what extent the 'Yes' group are relying on AI to actually do the bulk of their writing, or whether they're using it in the writing process, i.e. to do ideation or create briefs. If 37% are actually using AI-written stuff I think that's quite a high number.
What are AI content tools currently being used for?
The survey respondents who reported currently using AI tools were asked what they actually use them for.
Optimising existing content is the most popular use for AI (58.9%), although writing long form content (including blog posts) was close behind (57.1%).
Optimise existing content
Write long form copy, e.g. blog posts
Write short form copy, e.g. social media posts, product descriptions
It is surprising to see long form content so high up this list. Although that is the use-case that many of the tools have aligned themselves with, it is actually one of the weakest areas for the the current generation of tools (based largely on GPT3).
I’m about to dive into a project with the founder of Moonbeam, which is built specifically for long-form content. Its aim is to help people share their studies, research, and results more easily — kind of levelling the playing field for folks who might be brilliant researchers but who dislike writing.
I’ve also played around with a few AI writing tools and a few of the “This X Does Not Exist” generators at thisxdoesnotexist.com. I think there’s huge opportunity here!
The top answer here — optimising existing content — was what I expected, but I was very surprised to see that over half of respondents currently using AI tools were using it to create long form content.
From what I've seen, AI-generated content isn't currently of a high enough quality to be used without significant editing.
AI content tools have been strategically positioned over the past few years as helpers with the on-page process, and only have recently several tools made an active effort in diversifying their tools with content rephrasing and content writing. Previously, and most commonly, such tools have been used for keyword suggestions, pattern recognition, title enhancements, and writing meta descriptions or summaries.
Optimising existing content is a great use of AI and one of its strengths, in my view. It's a good way to make quick improvements, such as adding semantically-related keywords, and to spot gaps in the copy when compared to competitor content - which you can all do in one platform with tools like Frase.
We also use it in our copy brief creation process, though we use it for guidance on what to include, rather than relying on it to automatically build us a brief. I still think the human touch is needed to ensure a brief leads to quality work that's right for each client.
It's really surprising that using AI for long-form content is more common than for shorter copy or brainstorming. Current AI is much better at ideation and short-form creation than building a full, cohesive article.
Do you use any of the following tools for content creation?
The survey respondents reporting they currently use AI tools were asked about the apps they rely on.
Jasper is by far the most popular AI tool, with 43.9% of those reporting that they use AI tools using it. However, it's also worth noting that 'Other' was the joint-most popular option, suggesting there's a wide range of AI tools currently being used.
The majority of these tools are effectively interfaces for OpenAI. Some add additional training into the model, but many are simply a UI layer over the same tool.
Frase is my favourite currently, mostly for optimising existing copy and helping us create comprehensive copy briefs, as well as providing writers a bunch of research in the form of ranking pages in one place.
How much do you pay per word?
We wanted to know which pricing models the survey respondents are currently seeing from AI content tools, whether that's price per word (and how much they were paying per word) or a flat monthly fee.
The majority of respondents either weren’t sure how much they pay per word, or pay a flat monthly fee, which suggests most AI tools are not following a pay-per-word model.
Most AI tools are using a SaaS model, with a monthly fee for service access.
I expect that in the future, these tools will move to a pay-per-word model for more advanced services, where they might offer service users the option for a more advanced, custom-trained model for generating their content. They can then capture niche industry users with more specific content needs that might not be readily served with the standard models, typically used out of the box.
My team doesn't rely on AI to do any actual writing. I've tested this out quite extensively and the quality is not currently good enough.
I really hope that AI tools can find a better pricing model than pay-per-word. The real problem here is that AI takes iteration and practice to get good, meaning you have to use it a bunch. Capping people on words will incentivise them to use it less. I think a pay-per-word model shoots these companies in the foot.
If you’re not currently using AI tools, have you ever used them in the past?
The 63% of survey respondents that reported not currently using AI tools were asked if they'd ever used them in the past.
Less than a quarter had any experience of the technology.
Why aren’t you currently using AI-generated content?
For those not currently using AI tools, we wanted to know the reasons behind it. Respondents were able to choose multiple options.
Concerns over quality was by far the most common reason for respondents not currently using AI tools, but not knowing enough about it was also cited as a key reason.
I don’t believe AI-generated content is of a high enough quality
I’ve used it before and didn’t get the results I wanted
Our clients would be unhappy if we used AI to generate their content
It’s not cost effective
Survey respondents also shared with us other reasons for not currently using AI-generated content.
"I need to learn more about how to do it and what tools to use."
"Clients don't always know what they want, or focus on the wrong messages. It takes a skilled writer with an audience focus to create engaging content."
"Machine content has no place in the writing industry."
I think this is a bit of a scenario where people are still considering AI to be a replacement tool rather than an augmentation tool. You don't rely on AI to create high-quality content. You rely on AI to help you create a piece of content that is better than what you originally would have developed had AI not existed. The shift in your mindset will give you the ability to create pieces that are more likely to achieve your goals and objectives.
Let's imagine that steroids used in sports were okay. If you have two athletes competing and all things are created equal, except one takes steroids and the other one doesn't, the athlete who takes steroids is going to perform better. In many ways, AI can be considered steroids—except they're legal. They can help you perform better and they can help you get a better result. But for some reason, we continue to treat AI as if it's something we should be afraid of rather than something we should embrace.
It's worth noting here that this was a multiple choice question.
While not believing AI-generated content is of a high enough quality received 69.1% of the total responses, 86.6% of people asked this question gave this reason. Clearly, concerns about quality play a huge role in holding brands and agencies back from using it more.
However, it's also interesting that about 10% used the 'Other' option to report that they either didn't know enough about AI content tools, or didn't know where to start. There's clearly some education AI vendors still need to do here.
I think that people who think the quality isn't good enough are actually saying they don't really understand how to use AI tools. The quality is good enough, but you do need some trial-and-error to fine-tune it.
From my experience of trying out a few tools, writing various types of copy for various clients/sectors, quality is nowhere near good enough.
I feel that clients wouldn't be very happy with us using AI either, unless it could perform better than humans and help deliver better results for them - which it can't.
I don't think most, myself included, have grasped the positive impact that AI can have on research, outlines, optimization, etc. We mostly focus on the creative. ("AI will never replace me!")