AI content has arrived…

Even a few months ago while attending the BrightonSEO conference, I was intrigued to note that much of the buzz concerned artificial intelligence (AI) content, and the potential opportunities and pitfalls that computer-generated content will have on our industry. Then at this last year's Web Summit in Lisbon the conference was packed with start ups exploring solutions in the AI content. But is their product – market fit? What does AI content mean for SEO?

Let's explore the impact, challenges and opportunities that artificial intelligence content provides.

Matt Bennett's excellent talk on the topic of AI, at BrightonSEO, really challenged the audience and posed more questions than it answered. It was also wonderful to meet John Mueller, from Google's Central Search Team – effectively, the man at the front-line of Google's response to automatically-generated content.

AI is coming on in leaps and bounds, and can play a huge part in SEO strategies by discovering opportunities, like related keywords. Its algorithms and speed can help businesses expedite processes and enhance the accuracy of keyword research, competitor analysis and search intent research, among other functions. But as we will see, while AI content has made huge strides, it's not quite there yet in a fully functional capacity. 

The state of AI content creation today

Agency Collective recently surveyed 64 agencies involved with content generation, of which 65% said they were already working with AI tools in some capacity, while 29% said they've used it in a production project.

The current leader in AI content generation is GPT-3, an autoregressive language model that uses deep learning and text scraped from the internet to produce human-like text based on 175 billion parameters. As you can see from the example below, it is capable of taking regular prompts like "once upon a time" and expanding them into fully-formed, coherent content, which can often be indistinguishable from text produced by human authors.

The all-important question for anyone considering using AI in a content generation and SEO strategy is: will Google allow it?  In August last year, Google announce the launch of a new algorithm update, called the Helpful Content Update. This update introduces a new ranking signal that will negatively impact sites that publish high amounts of content with little value, are low-added value, or are unhelpful to searchers. It aims to "better ensure people see more original, helpful content written by people, for people, rather than content made primarily for search engine traffic" – including AI-produced content. Effectively, Google will tolerate AI-generated content to the extent that it is useful to users – which should be the goal of any user-facing company anyway. But will penalize it when its main goal is to manipulate search rankings.

John Mueller recently stated that "if you're using machine learning tools to generate your content… it's still automatically generated content and that means for us (Google) it's still against the webmaster guide." Google can't automatically detect content generated by AI and as such, can't penalize it with any consistency. However, that's not to say they won't in the future. Ultimately, if you don't know how to use the tool properly, you could end up generating low-quality, unreadable content that could get you penalized by Google. And even if Google lets it slide, you're likely to see a negative impact on your website's traffic and engagement if your AI-generated content fails to entertain the reader or provide value. So AI is not the silver bullet which some content generators may think it is.

As for the question of whether AI-generated content can be trusted, recent market insights provide food for thought. When asked "Would you trust website content that has been written by AI?", 82% of respondents in a major recent survey said "no". Lack of trust was especially pronounced when it came to health information (88%), news (92%) and film reviews (94%). So it's fair to say that people don't trust AI content when it's explicitly labelled as such – and they REALLY don't trust some of it.

But it actually fares well when compared against human-authored content in a blind test. In one such test, human-generated text rated slightly worse than AI-generated content in terms of trust (73% mean trust score compared to 74%), engagement (67% mean trust score compared to 69%) and knowledge (73% mean trust score compared to 77%). So it may be that the online users actually prefer reading some kinds of AI content over its cheap human-produced alternative.

Main AI content creation tools

•       Copysmith – Copysmith is an AI copywriting software used for generating marketing-related content. This content includes product descriptions, social media ads, Google ads, blog posts, and others. It is best suited for eCommerce marketing teams and large marketing agencies.

•       ChatGPT – ChatGPT is an chatbot on steroids. It has multiple uses many of which we don't even know yet. This bot can help people and companies to create quality content at a faster speed, including emails, business pitches, blog posts, Facebook and Twitter ads, landing page, product descriptions, etc. You name it, ChatGPT can do it.

•       Jasper AI – Jasper is an AI writing tool for creating and improving short-form and long-form content, including product descriptions, Facebook ad primary text, photo captions, YouTube video descriptions and more. It's the AI tool which I've used the most to date.

The advantages and limitations of AI content creation 

AI content is faster, cheaper and easier to produce than most content written by humans. But large limitations still exist. And it is often indistinguishable from human-generated content, from the point of view of your readers and Google too, which means your brand and SEO positioning may not suffer if it's used properly but nothing is guaranteed. It's always advisable to stay within Google's terms and conditions.

Google itself uses AI to help algorithms discern searchers' intent and rank pages accordingly. So AI is driving Google's goal of producing the best search results for users. This is good news for companies hoping to create the best possible content for a topic. But AI won't be helpful for those who don't know what to write about, what to emphasize, and how much.

Use-cases for AI content creation include:

•       Writing Ad Copy

•       Writing Product Descriptions for millions of products

•       Writing URL or meta descriptions

•       Helping with content ideas when brainstorming

•       Creating data-driven content

•       Predicting winning creative (e.g., digital ads, landing pages, CTAs) before launch and without A/B testing

•       Choosing keywords and topic clusters for content optimization

•       Optimizing website content for search engines

Having examined the advantages of using AI content, let's consider its limitations. By definition, AI content is a rehash of information that already exists online, so it struggles to incorporate new information and shifting contexts. AI will tend to auto-complete "the President of the United States" with "Donald Trump", to give one particularly glaring example. Another notorious example concerns a gardening article about a "creeper plant" spitting out irrelevant content about a "creep". While the AI community hailed GPT-3 as a breakthrough upon its release last year, the program is essentially still a "very fancy autocomplete" according to Daniel Leufer, an expert on artificial intelligence at digital rights group Access Now.

Meanwhile, although AI content generation has remained fairly uncontroversial so far, high-profile AI missteps occur in other areas on a regular basis, from Microsoft's AI chatbot being corrupted by Twitter trolls and subsequently withdrawn, to Amazon's recruitment AI being scrapped after it was revealed the engineers trained it using misogynistic bias.

Indeed, AI often perpetuates the worst biases of its human creators, and the consequences of inbuilt prejudices are becoming increasingly apparent as politicians and society embrace this kind of tech.  Earlier this year, AI-based facial recognition systems used by police departments in the United States failed to properly distinguish between black faces, and placed a photo of Michael B. Jordan, the African-American star of Marvel's "Black Panther" movie, on the wanted list for a mass shooting on Christmas Eve 2021.

It's unlikely that your company's content generation policy will lead to missteps of this magnitude. But do you really want to take that chance while the kinks of the system are still being ironed out, and cheap, human alternatives exist in abundance?

A profound impact on the future of content

In conclusion, despite its obvious advantages, AI is not yet sophisticated enough to be relied on entirely for content generation. Even if we ignore the potential for disaster discussed above, it's my opinion that AI does not make content interesting or in-depth enough to make readers stick around, and needs oversight from a communications professional – or even an entire editorial board – to keep its overall quality in line with the requisite standards.

I am not against AI, but I do believe that it has limitations at present which are likely to turn people off, and impact results and sales in certain areas. As in every other aspect of your business, laziness and overreliance on technology is guaranteed to hurt your brand and damage its perception among users. Having said that, AI is the future, and sooner or later it will change the fabric of copywriting and SEO as we know it.