Since GPT-2 first made waves – and headlines – in 2019 as an AI tool that produced such convincing writing, it might be “too dangerous to release to the public”, AI copywriting tools have become vastly more available and accessible.
Google “AI copywriting tool” and you’ll find close to a dozen listings (including paid listings) for different AI content writing tools, interspersed with listicles rounding up the “best AI copywriting tools in 2022”. Automated content tools have emerged as one solution to the problem of content creation at scale: how to populate websites with thousands of product descriptions or content pages in a cost-effective manner. More than that, many of these tools also advertise their suitability for creating marketing copy, particularly search-optimised copy, promising an easy way for businesses to improve their search presence no matter their size.
Once upon a time, it might have seemed like a transparently poor idea to commission a machine to produce search-optimised content – a one-way ticket to incomprehensible, keyword-stuffed blog posts and a black mark from Google. However, with recent advances in technology, AI-written content seems more attractive than ever – with commentators even speculating that AI copywriting tools could replace SEO writers in their role as creators of content designed specifically for search. AI copywriting tools promise to create content on any topic, targeting any keywords – with one tool, Scalenut, offering SEO insights alongside search-optimised content, and recently netting more than $3.1 million in a funding round led partly by a former Head of Google Search.
For businesses eager to take advantage of these promising new tools, however, there’s a big catch. One of Google’s most recent algorithm updates, the ‘Helpful Content Update’ (HCU for short) – the latest in a long line of updates aimed at rewarding genuinely valuable content, and penalising content created simply to rank well – seems to single out AI-written or automated content as ‘unhelpful’.
Google’s blog post specifies that the HCU, which began rolling out in August and finished in early September, is “part of a broader effort to ensure people see more original, helpful content written by people, for people, in search results”, (emphasis mine), and adds that sites that use “extensive automation to produce content on many topics” could find themselves in the crosshairs.
Even for those brands that aren’t looking to AI-generated copy specifically for a boost in search, incurring a Google penalty for their use is very much not the goal. So, is the age of AI copywriting over before it began? Are there scenarios where it makes sense to use AI copywriting tools, or is it universally a bad idea? What separates these tools from an SEO writer who is also commissioned to produce content that performs well in search?
I reached out to three experts to get their takes on these questions: Niki Grant, Head of Search at independent media agency The Kite Factory; Joe Comotto, Director of Search Experience at digital marketing agency Incubeta; and Lemuel Park, Chief Technology Officer at enterprise SEO and content marketing platform BrightEdge.
‘Helpfulness’ is nuanced
“One thing we do know for sure, regardless of any ranking or algorithm, is that marketers who are not thinking about the content they produce online will not reap the full benefits – especially in conversion and revenue,” says Lemuel Park, CTO at BrightEdge. “This is true for unhelpful content, duplicate and auto-generated content [produced] for SERP rankings, or traffic only.”
In other words, focusing purely on gaining search traffic to the exclusion of all else is counterproductive – businesses have to think about how people will be engaging with their content beyond a click on a search result, and to what end, otherwise they won’t gain any meaningful interactions – or revenue – from consumers beyond that.
“The Helpful Content update (HCU) is part of Google’s evolution to focus on the user and their experience online,” Park adds. “Of course, content and SEO have always been a part of that. However, it represents a final shift away from SEOs who over-optimise content, and from black hat SEO and content tactics aimed at rank and traffic.”
Niki Grant, Head of Search at The Kite Factory, also points out that ‘helpfulness’ is a nuanced and subjective idea – it doesn’t just mean AI writing that might seem convincingly human on an initial scan of the content. “Whilst AI outputs may appear ‘more human than ever’, there are limitations in the AI capability that means outputs will rarely be on par with that of a human,” she says. “Cultural nuance for example; AI won’t be aware of colloquialisms or turns of phrase relevant to different regions. AI cannot use analogies or metaphors, which for many of us bring content to life and aid us in understanding a given concept: one of the core elements of content being ‘helpful’.
“The challenge is that what a user finds ‘helpful’ is subjective and nuanced, based on the audience, the topic, the nature of the content, and what the ultimate aim of that content is,” Grant goes on. “An AI tool will not consider or address this nuance, which may result in brands creating a volume of ‘same-y’, bland, tone-deaf content, which could do more to hinder than help not only a site’s presence in the SERPs, but its reputation amongst consumers.”
Whilst AI outputs may appear ‘more human than ever’, there are limitations in the AI capability that means outputs will rarely be on par with that of a human.”
– Niki Grant, Head of Search at Kite Factory
Joe Comotto, Director of Search Experience at Incubeta, points out that even prior to the HCU, Google’s guidelines already specified automated content is a no-no. Its Webmaster Guidelines, for example, say that webmasters should avoid “Automatically generated content intended to manipulate search rankings”, while more in-depth guidance on automatically generated content specifies that “In cases where it’s intended to manipulate search rankings and not help users, Google may take actions on such content.”
“In truth, when it comes to automated content the outputs can only be as good as the inputs,” Comotto says. “So, when it comes to content that requires thought leadership or creativity, the AI will create content that falls short.
“Take for example product reviews: you can easily tell the difference between low quality recycled content compared to in-depth reviews. Through the Helpful Content Update, Google is trying to combat reviews that just send a user to an affiliate buyer as opposed to an insightful review of a product.” Google rolled out a Product Review algorithm update just before the HCU, and Google’s guidance on product reviews – including its stance on affiliate programs, which is that websites featuring too much affiliate content may suffer in search – is thought to be important to staying on the right side of the HCU.
Do AI tools have a place in copywriting?
Is there a role for AI copywriting tools in content production at all? I put this question to our experts, who agreed that there are situations in which these tools can be helpful – but with caveats.
“If a brand’s constraint is time or resource, then AI tools can be helpful in creating a ‘first draft’ which can be reviewed, amended and humanised by an actual person,” says Niki Grant. “It’s worth considering that if you are producing content around a topic in which you are a bona fide expert, the value of the resulting content will be much higher if the words and sentiment are coming from a human expert as opposed to a machine.”
Joe Comotto acknowledges that automation may have a place in content creation, “especially on large websites that produce hundreds of pages”, but notes that it depends on the type of content and the purpose it is intended to serve. “It’s about deciding when you should rely on AI and when you need that human input. Product descriptions are examples of when AI might be helpful, but if you were looking to generate content that required more creativity you would want a human to be behind that.”
Even in the realm of product descriptions, however, businesses may need to be cautious: Lemuel Park points to research conducted by BrightEdge and Oncrawl into web experiences and technical challenges across four industries: retail, banking, insurance and real estate. It found that retail had by far the highest percentage of duplicate content at 26%, versus 7% for real estate, 5% for banking, and 1% for insurance.
Park specifies that this was down to product descriptions: “In particular, we noticed product descriptions (especially from manufacturers) created too many duplicate experiences. Hence, we advised marketers to focus on improving and adding unique content to get better results – even before HCU rolled out.”
Image: BrightEdge & Oncrawl
BrightEdge have found separately that article and category pages on ecommerce websites are much more likely to gain click-through from search than product pages, with articles about products receiving almost twice as many click-throughs as the pages for those products. “This is why the role of the content marketer is on the rise—especially established, authoritative and experienced writers,” Park says. “In fields like ecommerce—and based on our findings above—subject matter experts may replace automated writing for product descriptions. Definitely for long-form content such as articles and category pages.”
When it comes to the role of AI in SEO, Park believes that AI is best suited to providing data insights that can form the basis for a content strategy. “Data-driven insights offer a whole new level of business insights to our organisations. This helps them to react quickly to market shifts and identify customer intent, [which] helps them with topic selection for content writers. It can even help with optimisation recommendations.”
AI writing tools can be helpful “if they are there to assist with grammar, improving the text, and writing style. They are a good guide to help with content as long as humans supervise and edit,” Park says. “However, you cannot beat quality content written by humans for humans, and I would never recommend letting machines generate full articles.”
Is SEO writing best carried out by humans or AI – or not at all?
Many brands and businesses employ a writer dedicated to creating content that is optimised for search engines, usually called an ‘SEO writer’ or ‘SEO copywriter’. The practice of SEO is all about making sure a website and its content are as visible in search as possible, and SEO writing combines this with content marketing to create content that will rank well from the outset, typically in order to raise a business’ profile and bring in new sales and leads, and/or to establish it as an authority on certain topics.
With the rise of accessible and affordable AI tools promising to do this at scale, however, are the SEO writer’s days numbered? And considering that Google’s HCU takes aim at content specifically created for search, will SEO writers struggle to achieve cut-through even if they can still write more effectively than an AI?
Our experts think not. “The real value of SEO copywriters is that they help you understand human elements such as what people are interested in, your search demands and your user needs,” says Comotto. “So in the same way that you might have social listening tools or customer surveys to help you better understand your audience, SEOs can become your tool to gather that information.” These insights are valuable across the whole business, not just in the practice of creating high-ranking content.
Comotto also points out that creating content that will perform well in search is about more than pure content creation – presentation, formatting and markup all play a role. “For example, you can’t have massive chunks of text. You need to have your article structured in a way that’s very scrollable and browsable, because most internet users tend to skim articles. So the structure of the content is almost as important as the actual written text, and this is where your SEOs can become really valuable.
“They can also help implement schema markup, which is particularly important in helping Google understand the expertise of your authors.”
“We all need to bear in mind that pleasing Google’s algorithm is the means, not the end,” adds Grant. “Google has parameters in place to encourage the creation of content that users will find helpful and informative. Without a hard and fast measure of ‘helpfulness’, Google has no option but to judge this based on the culmination of multiple elements, hence algorithm updates designed to curb undesirable behaviours from website owners.”
Google’s algorithm updates, even in the 2020s where singular algorithm “events” are less common than they once were, tend to lead SEOs and website owners to debate how best to please the new algorithm, but Grant points out that this mentality is the opposite of what should be done. “The existence of the algorithmic elements does not mean that content should be created to please only the machine.
“If brands create original, helpful, valuable content, it will likely resonate well with human users. If brands [only] create content that the algorithm would consider to be original, helpful and valuable, some factors may be neglected, resulting in a smaller impact than expected.” This is the case regardless of whether the content is human- or AI-created, but as Grant argued earlier, human writers are much more likely to be capable of true helpfulness and value than AI.
“Poorly-produced content will not work,” in the era of the Helpful Content Update, concludes Park. “Content creators must understand the customer and the category while bringing their creative spin to the mix. Quality and experienced content writers will be essential.”
His advice to ecommerce brands in particular is to focus on collaborating with internal and external partners even more – “even before helpful content, changing consumer habits require new creative content strategies.
“We are entering a new era of content-driven ecommerce—helpful content focused on providing humanised content that answers a need with knowledgeable advice across all critical moments in their online journey—supporting the highest quality shopping experience possible.”
Global brands can also plan ahead and account for the algorithm update before it reaches their non-English-language sites, as updates are always rolled out for search in English first. “So, if you have sites in other languages, look and learn.”