AI generated content – will robots replace writers?
I vividly remember the first time I came across a news story about AI generated content that made me sit up and take notice of the technology’s potential. On February 26 2021, as I sifted through the day’s news researching the latest content trends for B2B clients, something made me stop scrolling – apparently, a robot had just written a play, and reviews from world-famous critics were making the national headlines. As self-doubt suddenly crept in, I found myself wondering whether my role as a copywriter was becoming obsolete.
This date represents a milestone in natural language generation (NLG), with the world premiere of When a Robot Writes a Play – the first play written almost entirely by a computer. The play was originally written in English and then automatically translated into Czech. Incredibly, 90% of the English version was generated by GPT-2, an unsupervised, deep learning language model, with only 10% manually written or edited.
GPT-2
Generative Pre-trained Transformer 2 (GPT-2) is an open-source artificial intelligence created by OpenAI. It has the ability to translate text, answer questions, summarise passages and generate text output.
The quality of AI-generated content
I know what you’re probably thinking: if AI wrote an entire play, could it also handle my marketing content? After all, there are countless examples of AI-generated text that support customer service needs, from chatbots answering queries to smart assistants summarising financial or business data. So, why should content marketing be any different?
The answer is simple. Although the GPT-2 language model technically did write a play, the quality of the script was debatable. Responses were mixed, with reviewers praising its pioneering use of AI but criticising its similarity to existing plays, the simplistic structure of the script, occasional nonsensical bits, and general lack of originality.
How do AI models generate content?
To understand why automatically written content lacks the finesse of a creative mind, we need to consider how NLG works. NLG is achieved by training a statistical model through machine learning, using a wealth of pre-written text.
For simple data-to-text applications, such as weather forecasts or financial reports, this approach has proven quite successful. In fact, for this kind of application, AI-generated text has been deemed comparable or even superior to human-written text.
Will content writers become obsolete?
Not quite. While AI can write simple text with the right grammar, good syntax and a coherent structure, there are crucial aspects of content writing that only a human can handle.
Here are our top five reasons why good content writers are still set to be in high demand.
1. Engagement and expectations
AI models can’t - not yet, at least - understand an audience and its expectations, or know how to keep them engaged. As a content writer, one of my favourite ways to achieve this is to interview the client and let their voice shine through the copy. Personal commentary, real-world examples, character and humour… this is what brings a text to life and makes it a great read.
Input from thought leaders has another crucial advantage: it brings the conversation forward. Insight from experts add new perspectives to what is already online, creating truly informative content, rather than a summary of what is readily available. Don’t expect that from AI-generated content; however sophisticated the software is, it can’t bring a view or opinion to life.
2. Uniqueness
No matter the quality of AI generated content, at the end of the day it will always be a combination of existing text. This puts brands using AI generated content at a distinct disadvantage: with an average of 4.4 million blog posts published every day, creating fresh and unique content is crucial to cutting through the noise.
Most importantly, this lack of originality means your content might end up being too similar to what already exists online. This makes it impossible to highlight your value proposition and puts you at risk of unintentionally plagiarising someone else’s text.
3. Fact-checking
Fake news is so abundant and pernicious that a new word was coined to describe the declining influence of factual truth in shaping the public opinion: “post-truth”. “Post-truth” was named Oxford Dictionary Word of the Year in 2016 and the importance of this concept has increased ever since. But how does this relate to content marketing?
To build brand trust, it is essential to provide content that’s not only engaging, but also factually accurate. For this reason, trade bodies such as the Chartered Institute of Public Relations (CIPR) offer guidelines and resources for fact-checking and ethical practice. This ensures that content written and distributed on behalf of our clients contains information from reliable and clearly identifiable sources, as well as the client’s know-how.
Unfortunately, it’s nearly impossible to determine the source of AI-written content, which is why there is a shared consensus that it should always be disclosed as such.
4. SEO
What does Google think of AI-generated text? According to Google Search Advocate John Mueller, it “fall[s] into the category of automatically generated content [and] it’s essentially the same as if you’re just shuffling words around, or looking up synonyms [...] That means for [Google] it’s still against the Webmaster Guidelines. So we would consider that to be spam.” Mueller also points out the webspam team can take action if they spot content that has clearly been written by AI.
Also, don’t forget that Google’s recent Helpful Content Update aims to strengthen the search engine’s core proposition, by prioritising content that provides valuable information to users and penalising copy that is dense with keywords, but lacking depth, fluency and originality.
5. Postediting
Every writer knows their initial copy might be radically different from what will eventually be published. Content creation requires a constant back and forth among several parties – the client, the writer, and in the case of editorial content, the publication. Several rounds of revisions are standard practice in content writing, and ensure all parties are happy with the final product.
With AI-generated content, the amount of editing needed might be so overwhelming that it will significantly offset the time and cost saving benefits of automating the writing process.
My carefully considered human conclusion? While AI-generated content shows some promising advances, content writers won’t become obsolete any time soon.
Need help crafting engaging content that cuts through the noise? Reach out to the Marketing Pod’s content centre of excellence. Our lovely humans will be happy to help!