The tech behind NLG: what is GPT-3?
If you’ve heard about NLG, then you might well have heard of GPT-3. By way of definition, GPT-3 (which stands for Generative Pre-trained Transformer 3) is essentially an intelligent software designed by OpenAI to read and understand words and then generate copy in a way that can be understood by humans.
It can create anything that has a language structure, meaning it can write essays, summarise texts, answer questions and much, much more. A lot of NLG tools on the market including Jarvis and Copy.ai, utilise this software in order to produce copy, help create content ideas, and aid marketing efforts. In short, it’s the tech behind these tools that does the clever part.
With this in mind, and GPT-3 already being utilised by various NLG platforms, it begs the question, do we still need copywriters, and if so, why?
1. Because your content needs that human touch
First things first, it’s important to bear in mind that by and large most content is produced for and consumed by humans. Therefore, it’s safe to assume that since we are human, we can think, speak, and write like one. More specifically, strategic copywriters go one step further and take the time to fully understand the intent of the page in line with the business’s overall goals and complete user journey.
Whether it be a blog article, a product page or a campaign page, whenever a reader goes online, they subconsciously assess a piece of content they are viewing and quickly determine whether they want to carry on reading or make a purchase or not. If they do, it’s because it’s interesting, resonates and/or is useful, fulfilling their need or requirement. Good copywriters understand this, and that in order to sell a brand or product, you need to speak to human emotion and experience – and authentically at that.
For now, AI-powered NLG tools are still lacking in this vital empathy, emotional awareness, and at times, common sense.
2. Because a bot’s work still needs polish
As NLG tools are still in their infancy, a lot of the content produced (at least from what we’ve seen) needs polish. Sure, you can give the tool a decent brief including keywords, messaging and topics, but most of the time what’s produced just isn’t quite up to scratch and generally not good enough to deliver to a client.
That’s where a copywriter adds real value. Not only in terms of the usual editing for spelling, grammar, tone of voice and so on, but to check what’s been written actually matches the brief and the client’s desired outcome.
Other things that good copywriters will check during the QA process, include:
- Tone of voice and style/brand messaging – Is the copy aligned with the brand guidelines?
- Checking that copy is effectively optimised for SEO – Does it target the queries potential customers are searching for?
- Ensuring the copy reflects the intent of the page and includes the right CTA – Does it encourage people to take the desired actions?
- Checking whether the content is structurally sound – Are the flow and structure in a logical order?
- Asking whether the content makes sense – Will the reader be able to understand it with little effort?
- If there is a content strategy, does the copy align with that? – Have the recommendations of an editorial or content strategy been implemented?
Content briefing and quality assessment must sit with a human copywriter to maintain that quality.
3. Because NLG tools are not suited to certain types of content
As this topic continues to gain traction, it’s clear that at present, NLG tools just aren’t very good at producing certain types of content. Whilst they are fairly good at coming up with snappy ad copy, titles, subheadings and objective, fact-based content, they really aren’t so hot on opinion pieces, longform content, or persuasive writing.
We also noticed – through trialling various NLG tools – that automated language tools aren’t at the stage where they can fully nail a specific tone of voice or brand messaging. Many of our clients have markedly specific tones and values. They utilise slang, audience-specific language choices and humour; this is effective and relatable for customers because it’s been written by a human who really ‘gets’ these little language idiosyncrasies.
4. Because copywriters understand strategy
One of the most important reasons you’ll want to invest in a copywriter – and a good one at that – is because fundamentally, they’ll have a far better grasp of you as a client and what it is you’re trying to achieve from a strategic perspective.
Are you trying to improve your rankings? Improve the user experience of a page? Or capture customers specifically at the top of the funnel? A strategic copywriter should know this from the get-go. They’ll be involved from the start of working with a client, so they’ll have context behind the work which impacts the quality of the copy you’ll end up with, and ensures overall messaging is consistent throughout.
Strategic copywriters also understand content quality factors like the importance of E-A-T, user experience and social shareability factors, and can make copy recommendations that acknowledge these. We’re not convinced an NLG tool can go ‘beyond the words’ when it comes to creating content, and thus it’s unlikely you’ll see the same level of valuable recommendations.
5. Because the human mind is truly creative
One of the most beautiful things about the human mind is that each of us has unique thoughts and perspectives, and we all have the power to unlock our creativity in new and exciting ways. A good copywriter will be able to do just that.
It’s worth remembering that an NLG tool is quite literally generating language and content from ideas that already exist. At the moment, they appear to be taking content on a particular subject from the web, trying to understand it, and then essentially repurposing it. These tools aren’t creating anything genuinely new, or coming up with original thought in the way a human would. That’s not to say that most of the human-produced content we find online is completely original and unique, as that tends not to be the case. But there are ways we can make the content we create a good deal more valuable, and it involves a lot of the practices rooted in traditional journalism. Fundamentally, research. Which brings us to our next point…
6. Because copywriters can incorporate primary research
Much like traditional journalists, where possible, copywriters can and should incorporate original primary research into their content. This could be expert quotes, interview content, survey results, or any other type of research undertaken first-hand. The clever people over at Animalz call this ‘informational gain’. It’s about considering what you can add to the conversation around a topic to provide a different take on things.
As content strategists and copywriters, we shouldn’t just be looking to competitors to see how we can improve our client’s content. We should be looking to make our content stand out altogether, in a way that benefits the user. This is what will set us apart from automated NLG tools.
At this stage, as we’ve established, NLG’s can only churn out content based on what already exists online. They are unable to produce original research or thought.
Does this mean that once they’re commonplace, we’ll be heading towards a homogenisation of content and a complete lack of original thought and ideas? Will that be the new state of play?
Our Senior SEO Consultant Jose Hernando says:
“The rise of new and sophisticated NLG models like GPT-3 show us the huge potential that this technology can have for content creation. It seems that although these models are very impressive for short-form content, we are not yet at a point where coherent long-form content can be generated for a wide range of topics. However, I can see this changing in a relatively short period of time considering the number of resources that huge companies like Open-AI, Google or Microsoft are putting into it and how easily accessible they are with many of them being open-source or via relatively low-priced APIs.”
However, there is a flip side to AI NLG tools…
Despite the argument for retaining actual human copywriters – and we think it’s a good one – that’s not to say NLG automation tools have no place in our working worlds. In fact, we can see there are several instances in which utilising a tool like this could really benefit your workflow, particularly in relation to making efficiencies and therefore potentially cutting costs.
Where a tool like this can really hold court, is in producing snappy sports match reports, weather reports – or really any report whereby there’s plenty of data to go by. This software tends to be good at simultaneously reading and understanding large amounts of data and writing reports off the back of it. Content that doesn’t need to have a persuasive slant or be rooted in opinion works particularly well.
Some content areas an NLG tool might be advantageous include:
- Product descriptions
- Localising copy – for example, from British English to US English
- Corporate reports based on stats and figures
- Metadata (titles & descriptions)
- Short ad copy
- H1s / H2s/subheadings
These pieces of content would undoubtedly need the human touch when it comes to a final QA and sign off, but the ‘hard graft’ so to speak, could save your copywriter lots of time. Time which could be better spent on strategy, editing, QA and a smooth delivery.
Concluding thoughts
The inevitable rise of AI-powered NLG tools shouldn’t be a concern for copywriters or agencies, it should merely force us to take greater stock in the value of human words. We need to consider where it is that copywriters add real value, and where we can use these tools to help make efficiencies to allow human copywriters to do the job better.
It’s easy to say the majority of NLG tools just aren’t cutting in and can’t be relied upon to produce client-ready copy, but this will undoubtedly change in the future and we have to presume they’re only going to get better.
Copywriters will therefore need to focus on the added value they deliver, by futureproofing content and going above and beyond to prioritise informational gain, original thoughts, and content formats.
