Get the most out of AI as a writer

Don't blindly rely on artificial intelligence

Generative artificial intelligence (GenAI) is here to stay. Planning to do some writing? Then a little chit-chat with ChatGPT from Open AI, Copilot from Microsoft or Gemini from Google is almost part of your standard research routine. However, know that ChatGPT and Copilot, like other Large Language Models (LLMs), know absolutely nothing with certainty, even though they are learning fast. They only answer your questions with assumptions they present as truths. While the increment in data and computing power increases the probability of correct answers, the underlying methodology remains essentially the better guesswork. Should the phrase 'fake it 'til you make it' not already exist, they ought to invent it for GenAI.

Just to say: feel free to get GenAI on board, but don't rely blindly on it. GenAI may still be far from flawless, but in many aspects of the writing process it is a valuable new colleague. Master yourself in writing good prompts (see below) and check for yourself how such an LLM looks like under the bonnet. For example, if you know that this neural network predicts the logical next word every time based on billions of parameters, you will also better understand how to interpret the answers.

Anyway, always perform a fact check: is what ChatGPT is saying even true at all? And always ask yourself whether the answers generated are relevant and sufficiently sharp. Because even if they seem quite elegant at first glance, on closer look, they can be quite rattling. In other words, use your natural intelligence to filter out the artificial intelligence's quatsch.

Generative AI mainly reaps what we sowed

Artificial intelligence is the ability of a computer system to learn from its environment. AI learns from experience and adjusts itself based on feedback from the environment or the user. Reaping what we all sowed, in other words. Although AI does more than just rehash, the generated answers to your questions are little original for now, but largely a generic hodgepodge of existing texts.

Unfortunately, not all the information such a system feeds on is entirely accurate. The data used to train GenAI contains inaccuracies here and there. Unconsciously and intentional.

Unconsciously? Suppose Galileo Galilei had asked ChatGPT in the 16th century whether the earth revolves around the sun, the chatbot would have denied it. After all, most of Galilei's contemporaries did not know any better. And AI tools just tend to reinforce the consensus, whereas a critical writer is just supposed to question things.

Intentional? According to the World Economic Forum's (WEF) Global Risk Report 2024, disinformation will become the biggest risk to our global society in the next two years. With that, it knocks other lacklustre toppers like extreme weather events, cyber insecurity, armed conflict, inflation and social polarisation. WEF experts see two main reasons: the success of populist politicians ánd ... generative AI. Of course, a combination of both - populist politicians abusing generative AI to mislead people and bend them to their will - is extra dramatic. Especially when you know that fake news circulates up to six times faster than real news, according to a study by the Massachusetts Institute of Technology.

Inspiring sidekick, tireless sparring partner

Don’t think this will be a doom-and-gloom blog: generative AI also has a lot to offer. You can use AI for inspiration, for instance, to find a fresh approach to a topic and generate some fire starters. ChatGPT and consorts are ideal writer's block grinders. Even a suggested approach you disagree with can get you started.

Brainstorming also works well with GenAI, because it is never tired and always available. Furthermore, these tools are rarely short of cool anecdotes or one-liners to spice things up. They also often help you find a suitable logical structure for your text or story. You can also use a chatbot to check at the end of your writing process that you haven't forgotten a crucial element - such as an important advantage or disadvantage - in the text.

Perhaps the very best thing about GenAI tools is that they just speak English, or any other normal human language. Therefore, you don't have to learn some wacky programming language. Do you have a question you can put into words? Then you can ask it to ChatGPT & co!

Copywriter and copy editor

GenAI can help you write blogs, social media posts, web copy, email marketing campaigns, product descriptions, job postings, press releases... just about anything, really! And that in the capacity of both ''copywriter'' (from scratch) and ''copy editor''. You can give ChatGPT and consorts a great variety of roles.

Consultant: You do not pour your text-in-progress into the tool but ask for additional inputs that you may or may not include in your text.
Editor: Do you have a blog post that could use some refreshing? Copilot & co lets you improve existing texts. They suggest alternative sentence structures and synonyms. This is how you transform a boring paragraph into a more compelling story.
Tone-of-voice reviser: GenAI can also help you in your search for the right tone-of-voice. Want to transform a formal report into a light-hearted newsletter? That's possible! Or type in a sentence and ask the tool how it would sound, for example, if it were an Instagram post by a beauty influencer.
Corrector: AI can help you check your work for grammatical errors and typos.
SEO optimiser: AI can give you tips to make your text more findable online, although these tips are based on frequently occurring words. Effectively retrieving keywords from Google Site Console, for example, is not yet in the cards.
Secretary: Have an AI transcription tool - such as Otter.ai or TurboScribe - write out a first draft of your recorded interviews.
Test audience: Ask AI to predict how your text will be received by your target audience.
Sparring partner: As mentioned, brainstorming works quite well with GenAI, due to never getting tired and always available.
Translator: ChatGPT & co can translate your texts, but a dedicated translation engine like DeepL does a better job for now.
Networker: GenAI tools are well on their way to becoming good networkers. Not in the old school sense, but in the literal sense: you can link the tools with other sources. Soon you will link Gemini to your Google Ads suite, for example, giving you text suggestions that are more closely aligned with what really scores for your campaigns. This complementary, interwoven information can make your texts more targeted and therefore more impactful.
Primer: Slightly related to the role of networker is that of primer. As a master painter/writer, you can add depth to the flat primer that GenAI paints for you. With your own sources, feeds and api's, you add context that makes it more specific and increases quality. Learning to write with AI is learning to switch gears with data.
Writer: Finally, you can ask AI to write texts itself. But then you must give the bot a clear briefing and good prompts. How? You can find out below.

Priming, prompting and pimping

Are you giving a GenAI tool a writing assignment from scratch? Then be sure to provide proper context and a crystal-clear briefing. Such a briefing consists of priming and prompting.

During priming, you prepare the tool by entering relevant information or context. For instance, you can give the bot a role: "You are HR manager in a SME" or "You are a leadership coach". Context could be, for example: "Our insurance agency specialises in fire insurance. Personal contact is very important to us." You can also specify the tone of voice: "Make the occasional ironic joke" or even "Sound Obama-like".

The prompts make up the content briefing. Try to be as specific as possible: "Write a 1,000-word article on the benefits of system helmets for commuting motorcyclists in Belgium." Then you will refine the result with some additional questions: "Not bad, can you give me more advantages? And there could be a bit more humour in it!" When you are satisfied with the result, or feel you are running out of ideas, you start polishing the text yourself until you get it just right. This final stage is called pimping.

Basically, you work on a text together with your GenAI tool. Compare the collaboration to working with a new colleague: you must get to know each other; it doesn't run super smoothly from the first time. Rather, it is an iterative process, the goodwill must come from both sides. That iterative character is in fact leading for almost all types of output you create together with GenAI, not just for texts.

Knowing yourself is the beginning of all wisdom

As you know by now, generative AI (still) suffers from a lot of flaws and limitations. What is witty, though, is that you can simply ask AI tools about their own blind spots. For example, if you ask Copilot what the weaknesses of AI copywriting are, you get the following list:

Lack of originality and creativity: AI systems follow strict grammatical rules and structures, but often lack the creativity and originality that a human writer can provide. This can result in boring or styleless copy.
Limited contextual understanding: Although AI copywriting uses natural language processing, it does not always understand the deeper context. It may struggle to interpret ambiguities, humour, cultural nuances and specific jargon.
Overdependence on training data: AI models are trained based on existing texts. If the training data is limited or biased, this can lead to inaccurate or biased output.
Potential errors in (f)actual information: AI copywriting may contain factual errors if it does not have access to up-to-date and reliable information. It is important to verify the generated text before it is published.
Limited emotional intelligence: AI does not understand emotions and cannot show empathy. It can be difficult to write texts that have an emotional impact on the reader.

Already a good start, isn't it? Knowing yourself is the beginning of all wisdom. Copilot & co, however, suffer from even more flaws. For instance, they repeat themselves ad nauseam. Sometimes they also draw on sources from countries with laws, customs or traditions that do not apply to the text's target audience. In terms of grammar and spelling, however, Gen AI tools deserve kudos: this is often already surprisingly good.

Those who write will remain

Will AI take over the jobs of copywriters? Not immediately in too large numbers, we dare to hope. ChatGPT and consorts are not their competitors, but their assistants. If you want to enthuse or convince your audience, you still need sparkling, original copy. And for that, you need a good flesh-and-blood copywriter. Which is precisely the type of copywriter iO employs.